WO2002023899A2 - Method for the discontinuous regulation and transmission of the luminance and/or chrominance component in digital image signal transmission - Google Patents

Abstract

The invention relates to a method for regulating signal components of at least one data signal (DS) which is transmitted from a signal source (102) to a signal sink (111). The transmission takes place at discrete times according to whether and when there are changes in the data signal (DS) at the transmission end. A special example of embodiment of the invention comprises a method for automatic adjustment of the luminance, chrominance and/or colour saturation values in digital image transmission systems for video signals, for example in the television or multimedia fields. The invention is based on a method for transmitting differential images which are produced by subtraction of the YUV components of frames at scanning times which follow each other in time. According to the invention, these differential images are transmitted at irregular intervals at discrete points in time provided at least one suitably established time, Y, U or V threshold value is exceeded. This provides a means of considerably reducing the redundancy in the transmission signal and therefore, the data transmission rates compared to a periodically repeated differential image transmission.

Description

description

Process for the discontinuous control and transmission of the luminance and / or chrominance components in digital image signal transmission

A. Description of the general problem

The digital transmission of audio and video signals is becoming increasingly important today. Be it that compared to previous analogue transmissions with the same quality, more programs should be transmitted or a higher quality should be offered with the same number of channels, the solution is called digital transmission. Data is reduced first, because otherwise it is much too large

Bandwidth would have to be made available for the transmission. For this purpose, the source code is used to filter out the data that is irrelevant or redundant for the subjective acoustic or optical perception, so that there is practically no or only an acceptably small difference to the original signal during the reconstruction. However, this operation makes the signal more susceptible to interference. For this reason, appropriate error protection must be added for most transmission channels, which means that the amount of data increases again slightly. The problem of data reduction is, on the one hand, to keep the bandwidth B and data rate R of the transmission signal as small as possible, on the other hand, only as many signal components may be eliminated in the source coding by filtering that the transmitted signal can be received without audible or visible impairments ,

The subjectively not or hardly noticeable reduction of irrelevant or redundant signal components can be skilfully exploited by certain psycho-optical and psycho-acoustic effects can be achieved. Some of these effects are already being used in the area of source coding. In terms of optics, these include the following effects:

• Mach effect: contrast transitions on edges are perceived more intensely by the human eye.

• Oblique effect: Horizontal and vertical structures are predominantly found in nature. People are therefore less dependent on diagonal structures for orientation. The latter can therefore be treated rudimentary in the coding.

• Surface integration: The larger a single-colored surface is, the more precisely color and brightness values can be resolved - the eye can "integrate". The encoder must therefore quantize large areas (ie small spatial frequencies Ω _x or Ω ₂ in two-dimensional spatial space) more precisely than small ones.

Since acoustic or optical signals have a high proportion of irrelevant or redundant information, source coding can be used successfully with this type of signal. A signal contains redundant information if it can be predicted by knowledge of the statistics of the news source. If, for example, two neighboring pixels correlate with one another, one speaks of spatial redundancy; for correlated color areas or frequency bands, one speaks of spectral redundancy. The similarity of two successive images is referred to as current redundancy. The encoder can extract redundant information from the signal if it knows the source statistics. Apart from rounding errors, the signal can be fully reconfigured. be structured. The redundancy R (X) of a digital signal, which is represented by the bit sequence X: = [xι \ 0 <= i <= M- l} with the data bits xιe {θ; l}, does not therefore provide any additional information, but it is under certain circumstances advantageous for securing the useful information. To enable algorithms for error detection and / or error correction for a transmitted data signal, redundant information in the form of parity bits and checksums, for example, is necessary.

A measure of the information content of a digital signal, which is determined by the bit sequence X: = {i | 0 <= i <= M-l} is represented with the data bits xιe {θ; l}, the entropy H (X) of the signal. The entropy is a lower bound for the at best achievable average code word length. The Huffman algorithm comes closest to this barrier. The

The aim of the redundancy reduction is to reduce the data rate of the source by means of a transformation, so that an efficient entropy coding is possible. The signal components that the eye or hearing cannot absorb due to the limited resolution are irrelevant and need not be taken into account in the coding, but they cannot be restored by the decoder. Here too one encounters spatial, spectral and temporal irrelevance. Ideally, losses caused by data reduction can only be measured and not subjectively. In real applications, however, it is unfortunately sometimes unavoidable to do without parts of the relevant and non-redundant information if acceptable compression factors are to be achieved. The fundamental problem is therefore that at the moment it is not possible to operate a complete reduction of irrelevant or redundant data without having to neglect relevant and non-redundant signal components during signal transmission for optimal reception quality. The real-time digital transmission of audio and video data places unprecedented demands on the computing power of processors and the properties of the transmission networks, and various tasks have to be solved before a satisfactory solution becomes widely available. Digital video places by far the highest demands on transmission: it requires a large bandwidth, it must be transmitted with minimal delay and does not tolerate a high error rate. Even after compression and encoding, broadcast-quality video transmission still requires a data rate of R = 4 Mbit / s to 10 Mbit / s; this is an unacceptably high data rate for the local area networks (LANs) and most larger networks (WANs) available today. Delays in video data are critical because the individual images must be displayed with a refresh rate of _a = 25 (or 30) images / s. The real-time digital transmission of broadband signals therefore has the problem that, on the one hand, the networks available today cannot provide the required high data rate, but on the other hand, because of the desired high signal quality, the required minimum data rate must not be undercut, especially when transmitting digital image sequences in the multimedia field.

Real-time transmissions of video signals in particular do not tolerate fluctuations in the absolute delay time of the individual scanning images ("frames"), as they usually occur in most of the available networks, which is a further problem.

Communication networks are also never error-free and thus disrupt the transmission, for example by falsifying individual bits or losing entire packets. This in turn has a disruptive effect, since the compressed video signal is particularly susceptible to errors because the local and temporal redundancy was previously removed. Last but not least, the unavoidable interference in the communication network is the cause of further problems in digital real-time transmission.

The performance provided by the communication network depends on various parameters such as data rate, delay, delay fluctuations and error probability, which are referred to as "Quality of Service" parameters (QoS parameters). However, many of the existing networks do not guarantee the user specific limits for the QoS parameters, which means that these can change in an unpredictable manner. (An exception is e.g. B-ISDN over ATM, in which the QoS parameters are signaled when a connection is established.) In order to be able to offer a reliable and effective video communication service via various networks, it is important to know the QoS requirements of coded video signals. In the following, some important aspects in this context are highlighted, which show that it is not possible to meet all the requirements because some of them contradict each other.

First, the "ideal" transmission requirements for digital video should be considered, ie the level of quality that the digital networks should provide.

In an ideal digital video communication system, the image quality perceived by the user should neither depend on the encoder or decoder nor on the transmission medium. However, this is not possible, because in practice these two factors produce a signal deterioration that occurs in addition to the limitations of recording storage (recorder) and display (display, monitor). Assuming that such a system can control video quality on a specific ooo n- P- -3 H <JU P- ir rt P. SD -SP φ CD Hi Hi * J> τl ^► 0 <! P- CD H rt P Φ P CD LQ r

P Φ P 0 P- P- SD H SD c SD Φ P H- Φ M P = φ φ -: SD 0 3 Ω P. p. P P- φ rr P Φ

P LQ i-J i-J Φ φ -> SD t Ω tr H P. P tr P H tr tr K t P tr φ SD Pi P P Φ P P

• σ? F CQ n rt LQ φ tnm φ K. ω (- ^■ P Φ «P ¹ SD LQ φ LQ

Φ 0 CD H <rt Φ P f ω Ω -S 1-5 ω H Φ Φ 3 rr CD 0 φ P P. to CD to

Cl P- 3 NO Φ φ P iJ tr Ω tr H- LQ "\ HH SD φ rt 3 Ω Φ P P- tr ^• er P- rt φ P- 3 Φ tr ü LQ <i Φ t H CQ H SD t * ^ P o = 3 tr H LQ Φ Φ ra Φ SD P

P- Φ P H- rt N tu φ SD Φ Φ W Φ 0 P σi O φ H P rt -3 CD p. rr P Hi

Φ 3 ΓT N 0 = P- tr OHP H- Ω er iP & Hi Pi fi <rt Φ P Φ Φ P <Φ Φ φ Φ fö P- Φ LQ H ¹ φ (-li CD ω PP ^J P φ CD φ Φ H- Ω P P- ü P- φ P- P 3 P li

C P-? Ϊ ^* P. P- Φ P- D Φ C rt H rr CD φ P ill ω • P. i P CD rr P. rr N SD SD

Φ Ω φ rt H φ o H Ω SD h ¹ - φ rt P Φ _^ - _^ P - Φ SD P Φ N Φ Φ P SD P

P tT rr CD P. PH iJ tr tJ Ω H o = N ω P tr P <SD • 3 rr CQ Φ 0 P ra P- P CD er fT P- J iJ (-1- Pi Φ c-. P ^* P ι-S M- NP Φ P- Φ CD CD Λ P. rr rr Ω Φ

0 0 ω rt LQ SD Φ P- φ P Φ SD N -S H H α fd 0 P O P Φ SD SD = tr er

P- 0 3 ^■ <: Φ PH φ rt LQ PPP CD CQ SD SD P _^ - SD P er N o LQ CD iJ P- CD CD i P. Hl & 0. H- Ω ^■ -. ra P CD N <a ra Φ rt Φ P- t P? P. rt tr φ, -— -. rt CD ≤ ö P- -S SD P = Φ rt Φ Ω t- ⁱ P tr CD r P o * • p IS3 Φ -3 rt P- tr SD SD H- CD tr H h H tr ^• ~ -, rt 3 Ω * P- Φ li rr CD DPP

3 • 3 P- H Φ rt CD i CD Φ • φ Φ D IΛJ φ Φ Φ φ tr CD CQ i (D = -3 Φ SD CD CD

Φ w ω CD Φ φ i ^ P- Φ Pi H rt P- P o H- Ω 3_ P Φ r Φ Φ Φ rr P- P- tr rt o

H φ • P- φ i-J t? P. s! Φ Ω Φ P W CD P Φ P P- tr p. P tr SD -> - P- P- P. P iJ * J f H- 3 H- P tr CD φ rt - • Ω 3 P p, tT • SD = P iJ <CD P- φ Φ CD M Φ Φ H - 3 Φ tr SD Φ CD SD PP rt rt

P- Φ P- rt CQ -3 tr P. Ω H- CD P ^J ? F tr H ffi -> P. _^ -, P -SPPH LQ Φ

O P. P- P-! -R i-J H rt -S 2! D CD Φ P H- P N Hi P- φ P LQ P P- CD tr -3 φ tr rt Ω Φ; Φ Φ P- Φ tr H- H- P to Ω li P CQ Φ i LQ Φ α

Φ SD 0 φ SD i H rt H- <0 H CD H. P Φ Ω P LQ ~ —- tr Ω 3 H; rr <i SD -S P- P- tr rt CD l- ¹ rt P H-! -J H- 3 Φ H- p. PP tr N Φ Φ tr P 0 -— 0 P P- <P Φ

<^ Φ 0 φ P- - 0 φ P. Ω rt Φ P H td P Pi 3 Hi p. o ra σ.

Φ iJ P ¹ iJ 3 LQ iJ iJ Φ t • fl tr Φ <tr Φ <<φ P. P- SD? R 3 φ n Φ Φ P. Φ p. o Φ H CD H- Φ SD P H- SD H- Φ P- Φ tr 3 Pi ti tr Hi φ H P- H- <1 fr rt SD φ Φ ra H, Φ P LQ CD H CD <H tr 0 P- P- to PP =

0 = o H P- Φ 3 0 0 H 3 H- SD - ¹ P H- P - ¹ Xi 0 rt P: 3 φ Ω NW tr

CQ P CD P.] i-J 3 SD Φ P H Φ ω P Ct. Φ Φ o P- P Φ P Φ tr φ Φ Φ φ SD P- φ SD CD t) 3 Ω ω CD Φ P Cd φ H H φ Q P 3 x P ii H

H H Φ tr O t P- H- Ά s-; tr • Tj CD H- C H- H rr Φ • r. <! Φ rr O Φ φ P-

P Φ - Φ rt Ω φ iJ rt y φ • x] 0 l- ⁱ Φ P φ P- P SD 3 P- P Φ P.

P iJ • P. O tr P. t H- Φ H- tr (- ^■ s: Pi tu CD φ i tr Pi P φ rt P- 3 P-

IQ Φ i-J H? R H P H- Ω φ H- H- Φ s; P Φ Φ P P φ

<JHP Pi SD 3 φ SD -3 Φ 3 tr PH, -. H ¹ tr SD 0 H φ 0 P rr? tr CD

Hi Φ P <rt tr P o = H rt -S φ 4 3 Φ P Pi TJ Pi li tr P. Ω li CD SD SD SD o CQ 'ii Φ Φ P. Φ Hi LQ iJ H- SD LQ Φ ii ^ H- ii φ tr P Φ NPP Φ

53 ^* N SD iJ P- P- 0 tr 0 H Φ? SD P Φ PH Λ P Φ P rr CD

H 0 = p ⁱ P. φ 3 H- P. iJ ü SD P- P t? H Φ 3 Hl 3 Φ ω 0 P CD rt SD rr CQ φ P <φ Ω (- CD CQ P φ H- H- φ 0 ^• O t Hi P- P. Φ rr ö p. P Φ 3

Φ H t Ω φ t-r i-T H CD Φ ω H -S P tr P - - i Hl 3 Φ rr Φ P SD SD Φ P rt

N H 2 Φ Φ p, li CD Ω * <P - SD N Φ Φ 3 H- P tr Hi li N P rr li (-. Pi Φ

P Φ P- ra Φ rt tr CD 0 Φ tr Φ rt P SD SD P P φ Pi Φ P-

P P- SD ö H rt 3 H- H- H h- ¹ HP rt P- LQ CQ tr <! ? P CQ <! φ = ϊ *. LQ CO P. 3 SD φ? P- φ 3 CD P LQ Hi H- IQ Hi H- (- ^■ ra Φ Φ Φ Pi p. N 0 tr

0 φ Φ Φ rt H H- 3 Φ rt Φ φ Φ Hi H, rt 0 H- CD Φ i P- Φ SD φ 3 Pi φ

3 i-J H <CD P- Φ Φ P P tr Hi 0 H P Ω ^ Φ li P CD et P SD H

3 CD Φ rt φ H- CD Φ 03 o rt tτö H- tr ra P- P Φ Φ φ 3 li rr fu 0 N H φ t P

P p M -3 N l- ⁱ 0 tr = 0 e CD H- 3 Φ φ rr rt rt P- CD

P LQ 3 H φ P rt C SD φ φ P rt <? P 0 CQ ü tr P- 3 Φ SD

P- Mi φ 0 = N φ Si CD P Φ H- - H- P 3 rr Hi Φ Φ Φ SD P Φ l iV rr rr P- CQ PCH rt Φ SD P H- PMP Hl Φ 0 P ii H tr tr P iJ ii Φ Φ i- Φ tr -> P φ H- -> Φ SD rr Φ iJ IQ PP CD CQ P r- ¹ SD rr Φ r

tion difficulties. For example, normal conversation becomes very difficult when the delay exceeds Δt _v = 300 ms. If the sound additionally leads or lags the picture, the user loses the synchronization (eg the lip synchronization when talking). For many video services used or in the planning stage (video on de and, home shopping etc.), it is important that there is no noticeable delay between the control commands of the user and the reaction of the service.

In the case of one-way transmissions (e.g. digital satellite television), the total delay time is not so significant, so it is all the more important to ensure that there are no fluctuations in the absolute delay. Since every image and sound information must be presented to the user at a constant rate, any fluctuations must be compensated for before the display.

A video transmission system is responsible for ensuring that the video data passes from one end to the other. The QoS refers to the service level provided by the transport system to the video application and is characterized by the bandwidth, the error rate and the delay. Conceptually, the transport system can be divided into an underlying network and an end-to-end transport protocol. The latter improves the underlying network's QoS to meet the needs of the service. The complexity of this depends on the mismatch of the offered and desired QoS. If the network already meets the required conditions regarding QoS, it does not need another transport protocol. For example, within the Internet, TCP ("Transmission Control Protocol") is used as an end-to-end protocol in order to adapt the QoS provided by the network to that required by the applications. If resources for each sales reservation within a packet-switched network, it is possible to give guarantees for the maximum delay and minimum bandwidth (eg for B-ISDN over ATM).

The communication networks available today have limitations in QoS. The effects of these limits on the transmission of coded video sequences are briefly examined in more detail below.

The available data transfer rate has a significant impact on the quality of the encoded video data. Current video coding methods are usually lossy, which means that important data is sometimes lost during coding and compression. In any lossy video coding method, the average bit rate of the encoded data is generally proportional to the decoded display quality. With encoders or decoders for video signals which are based on the discrete cosine transformation (DCT), for example with MPEG1 and MPEG2, the quality loss occurs during quantization. A coarser quantization leads to a poorer picture quality after decoding, but at the same time, a higher compression takes place and a lower data rate results. Thus, the quantization resolution is a key parameter for controlling the video bit rate.

The communication networks encountered in practice falsify the data to be transmitted by changing or losing information. Even a relatively low error rate or very little data loss can have a serious impact on the decoded sequences. Compression algorithms remove much of the previously existing local and temporal redundancy in the signal, which leads to problems much more quickly if the encoded signal is corrupted (in Compared to falsification of the uncoded signal). The quantized DCT coefficients are encoded with variable run length codes (VLCs) within a scan frame of coded MPEG video data. The decoder does not know the length of the current VLC from the start; if a VLC is now corrupted, it is possible that the decoder will incorrectly decode an entire sub-sequence. In addition, some data (eg DCT coefficients and motion vectors in MPEG) are not encoded independently (I frames) but rather relative to previous (P frames) or even future (B frames) data. Thus, an error in an I frame also leads to subsequent errors in dependent P and B frames. For such cases, markings are provided in the coded bit stream at regular intervals, which enable the decoder to be synchronized correctly. Such a mistake can have an impact across a frame until the next mark is reached. Motion-compensated temporal prediction can cause the error to propagate both temporally and locally. Errors that affect an important point in the image or spread over several images are very obvious to a viewer. For this reason, an encoded video signal can only tolerate a very low error rate before the quality drops noticeably. An error probability of more than P _b = 10 ^"6 (ie an error on M = 10 ⁶ bits) can already mean a noticeable loss in quality.

B. Known solution to the general problem according to the current state of the art

The transmission of image and sound signals is already based in part on the psycho-optical or psycho-acoustic properties of humans. For example, in the field of image transmission technology, important parameters such as refresh rate and resolution are designed that the eye receives the most natural possible image impression. In the following, the most important aspects of the compression, transmission and decompression of image signals according to the current state of the art as well as the reception and processing of these signals in the eye of the beholder will be briefly discussed.

The eye recognizes the three primary colors red (R), green (G) and blue (B) with color-selective visual cells. That is why a color camera records these three colors separately. In general, a distinction is made between the three color elements of brightness (luminance), hue (chrominance) and color saturation. These elements also play an important role in digital image transmission and image processing technology. Luminance refers to the light energy that reaches the eye. A color with high energy leaves a bright impression, the same color with lower energy is felt darker. While chrominance defines the type of color and thus the wavelength of the radiation received, the color saturation provides information about the white value of a specific color. In the case of an absolutely pure color, i.e. radiation with only one wavelength (monochromatic light), the color saturation is 100%. With a pure color mixed halfway with white, it is only 50%.

The image or brightness information of a video signal is also referred to as a luminance or Y signal. The different brightness values of an image are converted into analog or digital voltage values that appear in the video signal as changes in amplitude. The faster the light-dark information of the pixels of a picture line changes, the smaller the coherent parts of the picture are of constant brightness, the higher the frequency of the signal voltage. Since high frequencies are difficult to transmit, the display of image details, the so-called image resolution, is limited. The CCIR standard specifies the highest transmitted signal frequency of a video signal with ^ x = 5 MHz. This defines the bandwidth and the maximum image resolution.

The color information of a video signal is expressed in the context of signal processing according to the PAL television standard by a brightness value (luminance), a color value (chrominance) and the color saturation. These three components must also be present when transmitting a color image. To generate a color image, a digital color image or video camera converts the colored light incident through the lens into three electrical signal voltages U _R , U _G and U _B. These three signal voltages correspond to the red, green and blue components of the image. These signal voltages are referred to as RGB signals. The components luminance, chrominance and color saturation are already included in the RGB signal. The luminance signal is obtained according to the three-color theory from the RGB signal values, that is to say by additive superimposition of color signal components of the three primary colors of the additive color system. It is known from three-color theory that any visible color can be put together from the three primary colors by mixing colors. The primary colors are those colors that cannot be created by mixing other primary colors. According to a decision by the International Commission for Illumination (IBK), also called "Commission Internationale de l'Eclairage" (CIE), the primary colors red (R, λ = 700 nm), green (G, λ = 546.1 nm) and blue (B, λ = 435.8 nm). Additive color mixing of the three primary colors enables light of the most varied color to be generated. For example, to obtain the color white with 100% saturation, the signal voltages of the three primary colors are added using the following formula: ÜY = 0, 3 0 - UR + 0, 59 - U _G + 0, 11 - U _B

with the voltage amplitudes

Ü: for the signal voltage of the perceived color white,

U _R : for the signal voltage of the transferred color component red,

UQ -. for the signal voltage of the transferred color component green,

U _B : for the signal voltage of the transferred color component blue.

The information for chrominance and color saturation is obtained by subtracting the signal voltage uj for the luminance value Y from the signal voltages U _{R l} U _G and U _{B of} the RGB signal. The color difference signals U _R -γ, U _G .γ and Ü- _B - thus arise. Since the human eye is less sensitive to color differences in small image details than to brightness differences in these image sections and the broadband Y signal is transmitted separately, the bandwidth B of the color difference signal can be limited to β = 1.2 MHz.

The three components red, green and blue are sufficient for image reproduction in the monitor of the receiver. It is therefore sufficient (if the Y signal is present) to transmit only two color values and the luminance value. In practice, therefore, only the two color difference signals RY and BY are transmitted in addition to the Y signal. The color difference signal GY is not transmitted because it has the largest signal amplitude and therefore most influences the signal-to-noise ratio of the transmitted video signal. In order to further avoid overmodulation of the transmitter, the two color difference signals RY and BY are limited in their amplitude. The limited R-Y signal is then also referred to as the V component, the limited B-Y signal as the U component. The three primary colors red (R), green (G) and blue (B) can be recovered in the monitor of the receiver by adding the received Y component to the received one Color difference signals RY and BY added, i.e. executes the following matrix-vector multiplication:

Aι, oo 0.00 +1.00 0.00 +1.00

+ 1.00 -0.19 +1.00 -0.19 -0.51 BY

^ß y + 1.00 +1.00

+ 1.00 +1.00 +0.00 _j RY

The color green (G) results, for example, as

G = Y-0.19- (BY) -0.51- (RY) = 1.70 • Y-0.19 ^• B-0.51 -R.

European PAL color television represents a further development of the American NTSC system. In both systems, the color information is transmitted with the aid of quadrature modulation, since only one high-frequency carrier signal is available due to the system, but two color information signals (RY and BY for the components V and U ) must be transmitted with the same carrier. The color signal therefore consists of two modulated carrier vibrations that are 90 ° out of phase. In the NTSC system, however, the V component is only modulated on the positive V axis. There is no alternating switching of the ink carrier phase to compensate for phase errors in this system. As a result, the NTSC system is more susceptible to phase errors in the transmission.

In order to create a smooth image sequence, at least 16 full images per second must be scanned and transmitted. However, in order to obtain a flicker-free impression, a refresh rate of at least f _a = 50 Hz is necessary. However, it is sufficient to transmit two interlaced fields with 50 Hz each, whereby the bandwidth remains the same as with 25 frames. The number of lines (for PAL) was set at 625 lines so that the vertical resolution is approximately corresponds to that of the eye. The horizontal resolution is determined by the video bandwidth and has been adapted to the vertical.

Instead of a complete transfer of the three primary colors, methods are used that use less bandwidth. The brightness (luminance or Y) with the full bandwidth B = 5 MHz, the difference signals blue minus luminance (BY) and red minus luminance (RY) with the reduced bandwidth of approx. B = 1.3 to 1 each , 5 MHz is sent. The luminance is decisive for the image sharpness (full bandwidth), while the color information has to be transmitted less precisely. In the baseband spectrum of the composite signal transmitted from the Y, R-Y and B-Y signals, the luminance (Y) occupies almost the entire channel bandwidth. The color information B-Y (U) and R-Y (V) are modulated onto a carrier which - in order to avoid interference, save bandwidth and be compatible with the previous black and white television system - lies in the range of low spectral energy density of the luminance. The audio signal is also modulated and placed at the top of the channel spectrum.

Since the image and sound signals are still analog, they must first be digitized for further digital processing. The process of digitizing is understood to mean the three steps of sampling (in compliance with the Shannon theorem to avoid aliasing effects), quantization and coding.

Transmission channels and storage media have only limited capacities, which makes data reduction indispensable for technical and economic reasons. For example, a video signal has a data rate of R = 124.5 Mbit / s while a satellite channel with a bandwidth of B = 33 MHz with a bandwidth efficiency of η = 1.57 bit / (s -Hz) only one data rate

R - = B - η «52 Mbit / s can transmit.

After the analog-digital conversion, the analog RGB components are converted into the YUV color space by a linear matrix transformation, which reduces the spectral redundancy. Furthermore, the high frequencies are concentrated in the luminance component because the eye cannot resolve high chrominance frequencies so well. After low-pass filtering (to avoid aliasing effects), the chrominance components are subsampled horizontally and vertically by a factor of 2.

Two images of a video signal that follow one another in time are very similar - with the exception of scene changes - and contain redundancy over time. However, if only the difference image is encoded, the entropy is much cheaper. Further improvements can be achieved if not the difference image, but the motion-compensated residual and the corresponding motion vectors are encoded: Here, a macro-block analysis is carried out in which direction the content of a block shifts. The decoder receives this information in the form of motion vectors. For image reconstruction, the blocks of the old image are shifted according to the vectors and the residual is added. There are various methods of motion estimation. In the so-called "block matching" method, the macroblocks are shifted in a defined search area and a correlation factor p is determined. The accuracy of the algorithm determines the entropy gain and thus the data rate.

With the help of the discrete cosine transformation (DCT), spatial redundancy and irrelevance can be removed from the signal. The representation of the image data by DCT coefficients has the

The advantage that the spatial redundancy (e.g. colored areas) that can only be detected in the block network is reduced by the transformation, because the signal energy is concentrated on a few coefficients. These are the coefficients of low frequencies and above all the fundamental frequency. With the help of the DCT, however, spatial irrelevance can also be determined and limited: due to the Mach and Obelique effect, the eye can resolve high frequencies, particularly diagonal ones, less precisely. The corresponding coefficients can therefore be roughly quantized.

When quantizing the DCT coefficients, most of the values - with the exception of a few in the vicinity of the DC signal component - become zero. By reading after the zig-zag

Zack method, the zeros are combined into a zero sequence. The run length coding combines the number of zeros and the first next non-zero coefficient into a pair of numbers. These pairs are then huffman-coded by taking their probability of occurrence into account when assigning the code word lengths. The combination of these two types of coding is very efficient with regard to the bit rate and approaches the entropy to an optimum.

A number of standards exist in connection with the compression and coding of moving image sequences, which will be briefly discussed below.

M-JPEG ("Motion JPEG") is not an actual standard, rather there are different methods of how the JPEG standard is applied to motion sequences. This is a pure intraframe coding, ie there are no cross-image algorithms for reducing the data rate. The compression is carried out by an 8x8 DCT with closing quantization and run length coding. M-JPEG is mainly used in the field of non-linear video editing and basically allows data rates from a few Mbit / s up to that of the uncompressed video signal. The image quality also varies accordingly: a data rate of approximately R = 1 Mbit / s corresponds approximately to VHS recordings, a data rate of R = 25 Mbit / s is almost studio quality.

Various techniques are used today to achieve the highest possible compression rate: First, an appropriate bandwidth must be selected for the signal to be encoded. The algorithm then uses block-based motion compensation to reduce redundancy over time. Motion compensation can be done in three ways: (a) predicting the current image from the previous image, (b) predicting the current image from a future image, or (c) interpolating two adjacent images. The prediction error (difference signal) is then compressed with the DCT by removing the spatial correlation and then quantizing it. In the end, the

Motion vectors are merged with the DCT information and encoded with a variable run length encoder. An optional error correction is then provided for the video data. MPEGl is mainly used for the storage of video files on CD and in the home computer area on CD-ROM. The data rate (image and sound) is R = 1.5 Mbit / s according to the standard, but there are numerous applications that also use larger data streams according to the MPEGl standard.

While MPEG1 is used in the multimedia field with the storage medium CD at low data rates of R <1.15 Mbit / s, the MPEG2 extension is used for the transmission of television signals using the interlaced method. MPEG2 provides data rates of up to R = 15 Mbit / s, whereby R = 6 Mbit / s is comparable to the quality of the PAL standard and R = 9 Mbit / s sub- It is hardly possible to distinguish jective from the original (studio quality, RGB) (visual transparency). It achieves its high compression factors through differential coding with motion estimation, a DCT with adaptive quantization and subsequent entropy coding.

The amount of data required for the transmission or storage of a digital signal with a given quantization error can be significantly reduced in many cases if the statistical relationships between neighboring samples of the digital signal are taken into account. A known method for data reduction of redundant and irrelevant digital image signals is the so-called "differential pulse code modulation" (DPCM). Compared to technically more complex alternative methods, the simple method of DPCM coding is very effective. In DPCM, the estimation error, ie the difference between the original signal and a prediction of a predictor obtained by linear filtering, is transmitted instead of the original signal. The degree of data reduction that can be achieved depends on the quality of the prediction filter. In the simplest case, the predictor is a runtime element whose delay time corresponds to the sampling period Δt of the digital signal; the prediction value is then the last sample value except for the quantization error. An improvement in the prediction can be achieved if transversal filters are used as predictors, which form the predicted value as a weighted sum over several preceding samples. The gain that can be achieved with DPCM transmission can be determined as follows: Given a Gaussian-distributed test component s (t) with a finite length and thus finite signal energy

E = js ² (t) dt <00, the standard deviation σ _s and the transmission power S _a = σ. In system theory, due to the above definition of the signal energy E, s (t) has the unit [s (t)] = VlV; in practice, however, the signal s (t) is a real signal voltage u (t) (in V) for te [tι; t ₂ ] across a resistor R (in Ω) or a real signal current i (t ) (in A), which during the time te [tχ; t ₂ l flows through this resistor R (in Ω). The signal energy is then

t,

E _el = - -] u ² (t) dt = R -] i (t) dt

R -. -

for te [tι; t ₂ ]. After sampling and quantizing this analog signal s (t) (according to the definition in the sense of system theory), the digital signal s (n) with the autocorrelation function is created

+ co ςε » _s (m): = s (-m) * s (m) = ∑s (n) - s (n + m).

In this case it can be shown that the quantization error power N _q in DPCM transmission compared to pure PCM transmission by a factor

can be lowered. Here is an example: The luminance component (Y) of television picture signals is sampled during a transmission according to the normal "pulse code modulation" (PCM) method with a sampling rate of f _a = 1 / Δt = 10 MHz quantized, the signal being encoded with k = 8 bits / sample. At this sampling frequency, the autocorrelation function has the value for immediately adjacent pixels

In the simplest form of DPCM coding, in which the difference signal values of two immediately adjacent sample values of the original signal are transmitted, the quantization error performance can be increased by a factor

"q, DPCM 2- (1-0.97) = 0.06, ie 10-log ₁₀ ^q ' ^DPCM * -12 dB" q, PCM N "q", P " _™ CM

can be reduced. Since the useful signal power 5 _a in DPCM is the same as in PCM, the signal-to-interference ratio 5 _a / N _q is also improved by 12 dB.

In addition to the standards mentioned, a number of other standards in the field of image coding are common according to the current state of the art.

According to the current state of the art, cameras and video cameras with automatic aperture control, in the English term "auto iris control" or "auto lens control" (ALC), are used to regulate the brightness of recorded images. In the case of an ambient brightness characterized by the time-varying luminous flux Φ (t) (in Im) and the time-varying illuminance E (t) (in lx = lm-m ^"2 ), the aperture can be changed by an adjusting mechanism and thus that on the image converter falling amount of light Q = \ φ (t) dt (in Ws or lm - s)

or the exposure

H = J E (t) dt (in Ws - m or lx - s)

to be influenced. The opening of the aperture therefore has a great influence on the quality of the reproduced images. The aperture must be constantly adapted to the particular shooting situation. However, a certain delay time Δt _V B is inevitable for this. The automatic iris control consists of an electronic-mechanical control circuit that controls the aperture so that the image converter still outputs a medium luminance level. The control signal for the actuator of the diaphragm is obtained from a feedback circuit which is driven by a DC voltage U _s ~ ü ^~ γ which is proportional to the brightness. Depending on the complexity of the control circuit, lighting differences between ΔE = 10 lx and AE = 100,000 lx can be compensated for with the ALC circuit. Special requirements are placed on the ALC circuit. For example, spontaneous changes in brightness or strong contrasts in brightness within an image area must be compensated for as quickly as possible and without control oscillations.

In order to prevent overloading due to excessively high luminance level U, so-called signal processing "white limiter circuits" are also integrated as chips in conventional video cameras. C. Inadequacies, effects and disadvantages of the known solution

The video cameras available on the market today carry out automatic brightness control and automatic white balance. Reduction of redundancy for the digital video signal to be transmitted, in the time domain by the sampled and quantized digital data signal

s (n) = V ne IN ₀ ,

ie s (n) e {0, ..., 0, s ₀ , s _l7 ..., s _{n ,.} , , , s _N , 0, ..., θ},

is given, a method for compression and coding is used on the transmission side and a method for decompression and decoding is used on the receiving side, in which n difference images are used as the useful signal at each current sampling time

Δsn ₊ m = s _{n + m} - s _n V ne IN with 0 <n ≤ Nm for m = 1

two time-sequential scan images are sent and - delayed by the time period Δt _ü required for the transmission between the signal source and the signal sink - are received at the scan time n + m. Instead of the sampling time t _n = n-Δt or t _m = JT.-Δt, n or m were abbreviated in the formulas above.

A discrete cosine transform (DCT) is generally used for coding and compression in connection with a run length or Huffman coding. The current coded difference images are transmitted at each sampling time n, regardless of whether and how many pixels of the video signal to be transmitted during a sampling period Δt (Δt: = _n -t _n _ι) between two immediately consecutive sampling times n and n + m experience a change in their luminance (Y), chrominance (Ü, V) and / or color saturation. However, this method is not suitable for digital image signal processing because it causes a high load or even overload of the encoder and, despite run length and Huffman coding, causes a correspondingly high data transfer rate. This problem does not only exist with video signals, but also with audio signals. Therefore, methods of source and channel coding are used today in the commercial transmission systems for picture and sound signals, with the help of which the data transmission rate can be reduced. Nevertheless, these methods are not a satisfactory solution to the problem of achieving the largest possible compression factors without impairing the quality of the image or audio signal received.

A further inadequacy is that the described method for transmitting differential signals at each sampling time n does not take into account the limited capabilities of a person's auditory or visual perception system for recording and processing acoustic or optical signals. Differential signals, the magnitude of which fall below a specific intensity difference ΔIJN or a specific frequency difference ΔfjN, cannot be perceived or processed by the human ear or face. Nevertheless, in the conventional transmission systems for video and audio signals, differential signals are also transmitted, which are superimposed by interference signals in the vicinity of the signal receiver and are therefore partially or completely covered and consequently cannot be perceived by the human ear or eye. It therefore makes no sense to send or send difference signals at every sampling time n received, since the human ear or the human sense of sight is not able to

a) intensity differences Δl or frequency differences Δf, for which ΔJ <ΔI _JN or Δf <Δf _TO applies, and

b) useful signal components ii-uz (f) which are completely hidden by interference signals in the vicinity of the signal source and / or signal sink, for which - _z (f) ≤ istör (f) applies,

perceive and process.

Another problem, which is described below, is still unsolved according to the current state of the art. If a sub-area of a picture scene experiences a change in brightness, for example when the video camera is briefly swiveled towards the sun or a strongly light-reflecting metallic object, this leads to a brief (usually a few scanning periods Δt) overexposure of the film, as an automatic Aperture control of an ALC device integrated in the camera can only take effect after the delay time Δt _V B (Δt _V B> Δt). The consequence of this is that individual amplifier stages are overdriven due to these signal peaks, and signal distortion occurs. After the above-mentioned ALC process has taken effect, however, only a brightness regulation can be carried out for the entire recorded image, but not for individual partial areas of the image. This means that when the automatic aperture control is used, all the pixels of a playback image always experience a change in brightness, although under certain circumstances only a small part of the recorded scene has become lighter or darker. When taking a short picture of an object or a person against a very bright background (for example in the case of a backlight shot), the depicted object or person appears extremely dark on the playback image in relation to the background.

D. Special task to be solved by the invention

The object of the present invention is therefore to improve the existing situation using technical means. The invention is therefore primarily devoted to the task of providing a reliable, improved method for regulating the compression and coding or decompression and decoding for the transmission of digital or analog data signals, for example for video signals. In this way, the proportion of redundant or irrelevant information can be reduced and thus the data transmission rate R can be reduced.

This object is achieved by a device with features according to claim 1. The solution according to the invention is not limited to applications in the field of video transmission technology. It can also be used successfully when compressing audio signals. Since both video and audio signals have a large proportion of irrelevant or redundant information, these signal components can be omitted during transmission without the interference of the received signal due to the lack of important picture or. Sound components is perceived.

E. Solution according to the invention for the special task according to the claims and advantages of this solution

Regarding claim 1:

The present invention relates to a method for regulating signal components of at least one data signal (DS). To simplify the spelling, only from be a data signal (DS), it being no longer differentiated between the data signal (DS) and signal components of this data signal (DS). The data signal (DS) can be transmitted, for example, from a signal source (102) to a signal sink (111). The signal source (102) can be, for example, a signal transmitter or a recording device for moving picture signals (television camera or video camera), and the signal sink (111) can be a receiving device for moving picture signals (monitor or display). The regulation can take place with the aid of at least one digital correction signal c (n) (KS) at discrete points in time, depending on whether and when significant changes of individual components of the data signal (DS) are present on the transmission side. With the help of this method it is possible to neglect short-term interference peaks or dips in individual components in the data signal (DS) during transmission. As a result, the data transmission rate R can be decisively reduced for the transmission path between signal transmitter (102) and signal receiver (111). The data signal (DS) can be either an analog signal s {t) or a digital signal s (n), which is obtained after sampling (while adhering to the Shannon theorem) and quantization of the analog signal s {t) , The invention can be used, for example, as part of a method for compression and coding, transmission, and for decoding and decompression of digital signals s (n). The signal components of the data signal (DS) can be transmitted discontinuously (that is, at discrete times at regular or irregular time intervals) as soon as new signal components are available on the transmission side. Possible application examples of the present invention are, for example, the compressed transmission or storage of the luminance (Y) and / or chrominance signal components (Ü, V) of a digital video signal s (n) (DS) in digital image transmission systems or the compressed transmission or storage the volume (L) and frequency components (f) of audio signals s (n) (DS), for example in the hi-fi, multimedia or telecommunications sector, in particular for UMTS applications. However, application examples are also conceivable in which the signal transmitter, transmission path and signal receiver are localized in one device, such as for example when storing the luminance or chrominance of a recorded image sequence in video cameras (102). In this case, the transmission signal (DS) is the digitized YUV signal, which is output at the output stage of the image sensor (102) after the RGB components have been converted into the YUV components. In this case, the transmission signal corresponds to the compressed and encoded YUV signal recorded on a storage medium, for example a video cassette. In this case, the received signal is the -RGB signal which is output on a screen or display (111) which can be integrated, for example, in the video camera (102) and which is obtained after conversion of the decoded and decompressed YC / y signal from the YUV to the RGB system.

Regarding claim 2:

To reduce the data transmission rate R on the transmission path between signal transmitter (102) and signal receiver (111), a method for generating, transmitting and processing a digital difference signal (DDS) can expediently be used.

As (n + m, m): = s (n + m) - s {n) V n, m e IN with 0 <n ≤ N-m

are used, which is obtained by subtracting the digital data signal s (n) (DS) from a digital data signal s {n + m) shifted by sampling times. The process associated with the transmission-side determination, transmission and reception-side evaluation of the difference signals Δs (n + -7., - τ?) (DDS) can be done for each sampling time n (with 0 ≤ n ≤ Nm) include the following steps: 1. At the sampling time n + m, the value of the signal change of at least one digital data signal s (n) (DS) is transmitted on the transmission side by forming the difference of the current sampling value s _{n + m} at the time n + m versus one at an earlier time Time n sample s _n formed on the transmission side is determined. 2. The difference signal value obtained

ΔS _{n +} : = Sn ₊ m ^~ S _n

is digitally encoded with a sign and amount at the sampling time n + m on the transmission side to form a bit sequence. 3. The bit sequence of the current coded difference signal value is transmitted to at least one receiver at the sampling time n + m. 4. After the transmission time Δtü has elapsed, the transmitted bit sequence for the transmission of the difference signal value Δs _{n + m} from the signal transmitter to the signal receiver is decoded on the receiving side at the sampling time n + m. 5. The signed amount of the change in the signal value of the currently received difference signal value Δs _{n + m} is added to the corresponding signal component of the data signal sample s _n output on the receiver side at the sampling time n at the sampling time n n for the reconstruction of the data signal sample s _{n + m} . 6. The calculated new sample

of the data signal as (DS) is output on the receiving side at the sampling time n + m.

The scope of protection of the invention also includes exemplary embodiments which contain only the transmitting or receiving branch alone, that is to say only steps 1 to 3 or only steps 4 to 6. Regarding claim 3:

With the help of at least one digital correction signal c (n) (KS) determined on the transmission side, the transmission of the difference signal Δs (n + -7., - τ?) (DDS) can be intervened in a targeted manner. These interventions can take place at discrete times, depending on whether and / or for how long there are changes in the data signal (DS) on the transmission side. In a special embodiment of the present invention, the correction signal c (n) (KS) is a constant which is determined as a mean luminance or chrominance value over all pixels of the digital image signal s (n) at a time n; in general, however, c (n) is a function f _{x of} the uncorrected transmission signal s (n). The corrected transmission signal s ^ _orx (n) results as a function f _{2 of} the original transmission signal s (n) and this correction signal c (n). In summary, the result is:

s _corr (n) = f ₂ (s (n), c (n)) = f ₂ (s (n), f _λ (s (n))).

In order to simplify the nomenclature, a distinction should no longer be made between the signals s (n) and s _kθrr (n) in the following, but only a transmission signal s (n). According to the invention, the digital difference signal (DDS) determined from s (n) is only transmitted if the data signal s {n) (DS) exceeds or exceeds at least one predefined point in time n at least one threshold value σ _ref suitable as a reaction level. below. This threshold value σ _ref can ideally be set so that the change Δs _{n + ra in} the signal s (n) determined at the time of sampling (n + m) can just be registered by any person. Regarding claim 4:

At least one of the threshold values used as the reaction level can be a predefined delay time period Δt _v (for example Δt _v = 5.0s), which may include several hundred sampling periods Δt. Waiting for this period of time before sending a new sample value can ensure that no short-term, interference-related signal peaks are transmitted in the differential signal As (n + m, m), which can lead to over-regulation of the receiving device. Thus, only long-term changes in a signal are transmitted, that is to say signal changes which remain at an approximately constant value at least for the duration Δt _v after a sample value s _{n has been sent} . Hence then

Δt _v >> Δt or k »1.

With the aid of this delay time, it can be waited until briefly persistent interference peaks in a recorded signal (for example in the case of a brief backlighting of an image object with a video camera) have subsided. In this way it can be avoided that, for example in the event of a brief overload or overexposure of individual frequency or image areas of a recorded audio or video signal, all frequency or image areas are reduced in intensity by using a control circuit.

Regarding claim 5: In at least one of the reaction levels used

Threshold values σ _re f can be, for example, the fixed rms value _ref / 2 of an amplitude 5 _ref , the frequency f _ref and / or the phase φ _{ref of} an analog sinusoidal Act reference signal s _ref (t) = S _re f-cos (2π- f _re ft + φ _re ). In this case, s _ref (t) can, for example, be a predetermined AC voltage signal Ure (t) = C7 _re fcos (2π- f _re ft + φref) with the effective value U _r e _f / / 2 or alternating current signal i _re f () = I _re f cos (2π- _ref t + φ _re f) with the RMS value J _ref / -v2. If the corresponding amplitude, frequency or phase parameters s _n , f _n or φ _{n of} a digital data signal s (n) exceed or fall below at least one of these threshold values, the transmission of a new differential signal value becomes at the sampling time n + m Δs _{n + m /} Δf _{n + m} or Δφ _{n + m} compared to the sampling time n released. This ensures that only the signal changes that can be perceived by a human being are transmitted, transmitted and received. If the digital data signal s (n) is a digital video or audio signal, the transmission of Δφ _{n + m can be} dispensed with, since the human perception system is unable to provide phase information φ (t) or To register and process phase changes Δφ (t) of analog optical or acoustic signals s (t).

Regarding claim 6:

A special embodiment of the invention is concerned with an application in the field of data transmission technology for video signals. It is based on the method described above for the transmission of difference images, which arise by subtracting the YUV components from the scan images s _{n + m} and s _n at the immediately or not immediately consecutive sampling times n and n + m. To do this, the signal transmitter and the signal receiver must have suitable technical components for recording (eg video camera), intermediate storage (eg video recorder and video cassette) and playback of video signals (eg screen or display). Furthermore, the signal transmitter with devices for Analog-to-digital conversion, compression and coding of analog video signals; the signal receiver should also be suitable for decoding, decompression and digital-analog conversion of digital video signals. The channel capacity C of the transmission channel, i.e. the product of bandwidth B and transmission bit rate R, should ideally be large enough to avoid so-called aliasing effects if the Shannon theorem is adhered to and to be able to transmit video signals in real time without any loss in playback quality.

Regarding claim 7:

At the fixed intensity and / or frequency of one

Reference signal s _ref (t) can be at least one luminance (Y _re f) and / or chrominance value (U _re f, V _re t) of a predetermined video signal with which the samples of the luminance (Y) or Chrominance [U, V) of a received digital video signal s (n) are compared. According to the invention, these difference images are transmitted discontinuously, provided that at least one suitably defined time, Y, U or V threshold value is exceeded. Image data are therefore only sent if the new scan image contains image areas with new information that is not irrelevant or non-redundant for the viewer. In this way, the irrelevance or redundancy contained in the transmission signal and thus the data transmission rate can be decisively reduced in comparison to a periodically repeated transmission of the difference image. A renewal of the current displayed image in the terminal of a receiver is only carried out for those image areas in which relevant changes in brightness, hue and / or color saturation are registered. The threshold values Y _re f, c7 _ref and V _ret can be determined from the brightness, hue and / or color saturation values of the pixels of the reference signal s _re f. Regarding claim 8:

In almost all systems that use a type of image coding with data reduction, the final receiver of the restored image signal (video signal or still image) is the human eye. It should therefore be clear beforehand that efforts must be made to include a model of the human visual system ("Human Visual System", HVS) in the processing chain. This is the only way to ensure that a) during the coding process, the bits are preferably assigned to those parts in the output signal that correspond to those structures in the image to which the eye is most sensitive, and that b) a psycho-optical quality measure associated therewith can be created.

This evaluation is necessary because under certain circumstances, human eyesight is unable to perceive changes in luminance or chrominance of a useful signal under certain circumstances, for example in the presence of broadband interference radiation of high intensity in the vicinity of the signal receiver. In the following, the most important insights gained from visual tests on a large number of test subjects and confirmed many times in this context are briefly discussed.

Standardized procedures that are internationally recognized are used for subjective vision tests today to ensure the comparability of different measurement results. Technical boundary conditions such as viewing distance, background lighting, the type of presentation of the test pictures and test sequences, the questions and the assessment procedure are clearly defined. Such procedures are laid down in the form of recommendations from a group of experts from the "International Telecommunications Union" (ITU). Today's technical development keeps making new demands. demands on the test procedures, so that the existing methods have to be modified and new procedures have to be developed.

The integration of human eyesight into the transmission chain from the signal transmitter to the signal receiver is important so that a reduction in the irrelevance information can be meaningfully adapted to the human sense of sight. To do this, it is necessary to know some important effects that play a crucial role in vision.

When seeing, the light incident through the pupil and lens into the interior of the human eye is imaged on the light-sensitive retina approx. 1 _LN = 17 mm away. The incident light intensity is determined by means of a photochemical reaction using two types of sensors, rods and cones. The chopsticks are more sensitive, but do not differentiate between colors. To do this, they still work with very little light (night vision or skoptic vision). The cones are responsible for stronger intensities (day vision or photopic vision), fine resolutions and color vision. For the latter process, they are divided into three different types, each with an absorption maximum at a different location within the visible spectrum (for red-orange, green and blue-violet). Together with two sensors, the system allows vision over an intensity or luminance range of 12 orders of magnitude (L = 10 ^"6 to 10 ⁵ cd / m ² ). This means that within the various stages of the perception process, an adjustment to the background intensity in a large area takes place.

As already mentioned, the eye is able to process an enormously large brightness range (approx. 12 powers of ten), from the intensity of the starlight to the pain limit (glare). However, since the pupil can only regulate its diameter in a ratio of 4: 1, a very extensive adjustment to the average scene brightness at the receptor level of visual perception must take place, because it is neither possible to adjust the excitation frequency for the transmission impulses on the nerve fibers by a factor 10 ⁹ - 10 ¹⁰ to vary, it is still necessary in practice. A much lower dynamic range is sufficient for daily use. The dynamic range of the eye (with regard to the brightness information) comprises approximately two powers of ten. Various test results confirm that for a given background brightness (for the adaptation of the eye), the number of distinguishable brightness levels is in the range from about 150 to 250. The important finding from the experiment described above is that the smallest perceptible difference in brightness does not appear as a constant value, but is dependent on the constant ratio of the ambient lighting. The visual system thus responds, at least at or near the threshold, to slight changes, which means that the perception of brightness depends much more on the ratio of object to ambient lighting than on the absolute value of the former. For the image processing technology, this means that an image disturbance (eg due to transmission errors or quantization noise) is noticed much more strongly in dark parts of the image than elsewhere. In the

In practice, due to the always present ambient lighting near the display, any deterioration (eg caused by noise) is perceived less strongly in dark areas than in lighter parts of the picture. There is therefore a much more complex relationship between perception directly at the visual brightness distinguishability threshold, in areas near the threshold and in more distant areas than one might suspect at first glance. It is obvious that the smaller the visual angle from which they are viewed, the more difficult it is to perceive fine structures. After all, with increasing complexity, small details can no longer be resolved separately. This solidifies the claim that the spatial frequency response of the eye drops to negligibly small values at high spatial frequencies (the number of complete vibrations per degree of a sinusoidal brightness pattern that are distinguished by the eye). Experiments have also shown that the sensitivity of the eye to areas of constant brightness (ie spatial frequency Ω _{X / 2} «0 m ^{" 1} ) is low. So that humans can see anything at all, the excitation must be in a considerable frequency range between them In areas with fine details that require a high resolution, the eye is more tolerant of deviations in the absolute brightness, so the amplitude frequency response of the eye is changed by spatial effects: on the one hand, by the influence of different receptor groups that affect background and If there is a simultaneous contrast between a targeted area and the surrounding area, the perceived brightness of the targeted area changes if the brightness is varied around this area. The targeted area appears all the darker, the lighter the H intergrund becomes. The amplification of this effect also depends on the existing lighting: If the contrast between the targeted area and the background is small, stronger overall lighting increases the perceived area brightness and the contrast decreases. Conversely, if the contrast is large, the area appears darker when the lighting is stronger. The second effect occurs with an abrupt change in brightness and is responsible for making the contrast appear sharper than it really is. Thus, an area with large constant brightness the perceived brightness of an adjacent area with lower (but also constant) brightness. The phenomenon responsible for these two effects is known as lateral inhibition and can be modeled in relation to the spatial frequency response of the eye as a high-pass filter, which has a significantly reduced sensitivity to areas with constant or slightly changing brightness and the targeted determination of sharp edges (for Example, the outline of an object).

To create psycho-optical measurement and evaluation models that can be used for the purpose of data-reducing image coding, in-depth knowledge of the optical signal processing of the eye to the nerve impulses in the brain is necessary. There are several for this task

Methods have been proposed that correlate more or less well with human perception. The most promising among them is the model of just noticeable differences ("Just Noticible Differences", JND). However, there are also differences here, depending on which and how many of the previously known psycho-optical properties of human vision are taken into account.

For example, the psycho-optical properties can be used to determine the threshold values for the smallest perceptible luminance (ΔYj or chrominance change {AU ^ _/ AV _J ) of the received video signal as a function of the lighting conditions in the surroundings of the image signal generator, which are characterized by (Y, U, V) _env of human eyesight with regard to the amplitude and / or frequency modulation threshold, the perceptible frequency group width Δf _G and the effects caused by the masking of useful radiation by interference radiation are evaluated. In this context, it is important to focus on the phenomenon of

To refer to "local masking" that is related to the Mach effect, but still deviates from it. The exploitation of this masking phenomenon is a successful example of the inclusion of properties of the HVS in image coding algorithms. In a series of eye tests, it was found in different test subjects that the brightness change threshold was sensitive to the presence of a nearby jump in brightness. In other words, the threshold value of a perceptible change in brightness (ΔY _JN ) is higher than it would be if the brightness jump (ΔY) were not present. That is why the jump is said to "mask" small differences in brightness in its surroundings. Another important feature of this masking effect is that it is a function of the gradient of the change in brightness against which it is measured. (In a typical test image, it is therefore dependent on the envisaged detail.) This fact can be used to define masking functions on the basis of weighted derivatives of the brightness distribution within an area comprising a few pixels. These can be used to optimize the quantization process of coding systems by prediction.

Over the limits of the visibility of luminance (ΔY _JN ), chrominance (Δ / JN _J ΔVJN) or color saturation fluctuations of a stationary received signal are the thresholds for sinusoidal amplitude and / or frequency modulation (AM or FM) for monochromatic or for broadband test radiation information. An amplitude or brightness modulation threshold is understood to mean the value of the amplitude modulation degree m for which the amplitude-modulated received signal begins to differ visibly in the case of sequential presentation. Analogously one understands a quenz- or chromaticity modulation threshold that value of

Frequency hubs Δf for which the frequency-modulated and unmodulated received signal begins to visibly differ in the case of sequential presentation. Modulated cosine gratings for luminance (Y), for chrominance (U, V) and for color saturation can be used as the basis for vision tests. When viewing the test images on a monitor, the individual modulation transfer function (MÜF) can be seen directly as a psycho-optical limit of visibility of the cosine grating depending on the distance between the monitor and the viewer. To determine the MTF for brightness or color modulated cosine gratings, a test person displays the points at which the cosine gratings can just be recognized by a test person and connects them with polygons. The middle limit, at which test persons can just barely recognize the lattice shape of the cosine-shaped modulation, marks an MTF. These measurement curves can be found for all color values that can be displayed on a monitor. All measurement curves have a clearly recognizable bandpass characteristic of the human face. For the MÜF -τ. (Ω _ι , Ω ₂ ) brightness-modulated grating, only the mixed color "achromatic" or the respective mean color value can be seen with low color saturation. Above certain spatial frequencies (Ω _X , Ω), the color saturation must be constantly increased in order to be able to recognize the grating. This effect is related to the diffraction limit of the optical image in the eye for high spatial frequencies Ωi or Ω ₂ .

Looking at the MTF for brightness-modulated grids in the

Compared to color-modulated grids, it can be seen that the global minimum of MÜF m (Ω. _Lr Ω ₂ ) for luminance is shifted to MÜF Δf (Ωι, Ω ₂ ) for chrominance to higher spatial frequencies (Ωι, Ω ₂ ) , All color-modulated git ter show a strictly monotonous increase in MÜF Δf (Ωι, Ω ₂ ), which is below Ω _1/2 = 1.5 line pairs / degree, depending on the respective color type. Paying attention to the different modulation transfer functions of the different color-modulated grids can enable targeted data reduction for different color transitions.

Regarding claim 9:

A special exemplary embodiment of the present invention can be used to carry out the white balance of video signals s (n), in which luminance (Y), chrominance (U, V) and / or color saturation of the moving images recorded by a video camera can be regulated depending on the lighting conditions in the vicinity of the video camera.

With image transmission according to the usual standards, falsifications in the reproduced luminance (Y) occur under certain conditions. Images that contain highly saturated colors and hard transitions from the colored image areas to achromatic image areas are reproduced in a particularly distorted manner. These errors also occur in digital image transmission according to MPEG2, in which luminance (Y) and chrominance (U, V) are quantized separately, since the separation of brightness, color type and color saturation information is not perfect.

With the aid of a compensation method for regulating the luminance, it is possible to correct the luminance defects caused by the coding of the chrominance signal components and thus to improve the image quality.

The color sensitivity of the eyes depends on the wavelength λ or the frequency f = c _n / λ which changes with the speed of light c _n = c / n in a medium with the refractive index n propagating radiation. C denotes the speed of light propagation in a vacuum; their value is approximately

c »3.0-10 ⁸ m / s.

The spectrum of light visible to the human eye is in a wavelength range between λ _m ι _n = 380 nm and λ _max = 780 nm. Light with a large wavelength (near λ _m ax) appears to the human eye as red, with a short one Wavelength (near λ _m i _n ) as violet.

White light contains all light waves occurring in the visible spectral range. This can be checked if, for example, white sunlight is resolved into its spectral components using an optical system (prism). The sunlight perceived as white consists of innumerable individual color components that are only put together to form an overall color impression in the eye of the beholder. However, the different visible spectral components of light are not perceived by the human eye with the same sensitivity. The sensitivity curve of the eye has a global maximum with green light and a strictly monotonically falling sensitivity towards violet and red radiation. If there are no electromagnetic rays, that is, if all wavelengths of the visible spectrum are absorbed by a body, the human eye perceives the color black.

The color impression perceived by the eye is not only determined by the wavelength of radiation, but also by the light and lighting conditions in the area. For example, the red color of an opposing tands different in the sun than in a darkened one

Room.

The sensitivity of the human face is influenced by the adaptation of the eyes to the existing lighting conditions in the area. In the so-called "bright adaptation", blue or red light emitted by a light source must emit significantly more energy E = h - f = h - c _n / λ in order to be perceived as bright as yellow or green light. The constant h is Planck's constant; it has the value h «6.62-10 ^{" 34} Ws ^2. The relative light sensitivity - usually denoted by V (λ) - has a global maximum at λ = 555 nm in the case of light adaptation, and a global maximum in the case of dark adaptation The ambient light therefore plays an important role in the recording of an object with a color camera. The color impression left by an object is also influenced by the ambient brightness, ie the color of an object can only be objectively achieved with constant intensity of the ambient light determine.

To compare colors, white light, which you will hardly find in nature, would be ideal. In order to be able to compare the colors of different light sources with one another, the radiation of a so-called "Planckian radiator" is used as the "reference light source". A Planck radiator is an absolutely black body with a degree of absorption of 100%, which emits photons after being heated. The spectral diagram of this light depends exclusively on the temperature up to which the black body is heated. For this reason we speak of the "color temperature" of light. The color temperature of the light is identical to the heat temperature of the black body, and it is called the absolute temperature T _x in Kelvin (K) specified. Even at a relatively low temperature of Tι «1,000 K, the emitted light appears reddish. With increasing temperature, the color of the light changes from red (Ti »1,000 K) to orange, yellow (Ti» 3,000 K) and white (T _λ «3,200 K) to blue (T _x « 5,000 K). The maximum of the spectral energy distribution shifts towards shorter wavelengths with increasing temperature.

The color temperature of a light source is determined by comparison with the light of the heated black body. If, for example, the black body is heated to a temperature of Ti «3,200 K and the light from this body corresponds to the light of a radiant incandescent lamp, the light emitted by the filament of the lamp also has a color temperature of 2ι« 3,200 K. The comparison of different

However, light sources with the help of the color temperature are only possible with thermal light sources (candles, incandescent lamps, daylight, light from radiation-emitting celestial bodies, etc.) for photon radiation in the visible frequency range. The light from luminescent lamps (gas discharge lamps, fluorescent tubes, light-emitting diodes etc.) does not show a smooth spectral distribution. Light of this kind cannot therefore be expressed by the color temperature.

If you look at an object alternately with different light sources, whose color locations in the IBK color triangle are fairly close to one another, but whose spectral distribution differs greatly (for example, natural daylight with a cloudy sky and light from a daylight fluorescent lamp), it can happen that the object appears in a completely different color. It can be concluded that the human eye is unexpectedly sensitive to the spectral distribution of light sources. On the other hand, the eye hardly recognizes a color difference when the Differentiate color locations of the ambient light, for example when viewing an object with daylight or in the evening with incandescent lighting. The reasons for this lie in the ability of the human eye to adapt to the different lighting and lighting conditions in the environment (color memory). Here is an example: An open-air photograph of a white object with a color temperature of daylight of T _x = 6,500 K should be done in such a way that the recorded object also appears white on a monitor. For this purpose, the camera is set so that the color components red (R), green (G) and blue (B) deliver signals with the same amplitude (U _R = C7 _G = Ü _B ), that is, the set color temperature of the camera becomes T ₂ set to T ₂ = 6,500 K. When the same object was recorded - but now carried out in the studio with a white light-emitting photo lamp with the color temperature T _x = 3,200 K and the camera setting unchanged at the color temperature T ₂ = 6,500 K - the recorded object appears yellow on the screen. The recorded white object now has a color temperature of T _x «3,200 K, which means a shift in the color impression towards the red spectrum. The cameraman, on the other hand, always sees the object glowing white because the human brain has a color memory. The object shown on the monitor only takes on a white color again when the camera is readjusted to the ambient light.

In order to ensure that a white object - recorded at different color temperatures of the ambient light - always appears white on the monitor, the camera must be able to compensate for the color temperature. Such compensation is referred to as white balance or white balance adjustment. When illuminating a white body with a light source with a color temperature of T _x = 6,740 K, the three output signals C7 _R , u _G and U _{B of} the camera must be used be the same size. In practice, the white balance can be one

Camera can be done by three different methods:

• by an optical-mechanical process (compensation filter),

By an electronic process (control circuit),

• by a combination of the two methods mentioned.

As a rule, modern video cameras work best at a color temperature of T _x = 3,200 K. With this lighting, no white balance is required. Compensation of the color temperature with color filters is carried out in practice with the help of a switch for outdoor or indoor photography ("in-door" / "out-door" switch). For this purpose, a color filter is mechanically connected in front of the image sensor.

In connection with color temperature compensation filters, the color temperature T _{λ of} a light source is sometimes also expressed by the calculation variable M _x ("Micro Reciprocal Degrees", MIRED) with the unit μrd, which is the reciprocal of the color temperature multiplied by 10 ⁶ , results:

M ₂ [μrd] = • 10

T [K]

The following correction formula is used to adjust the color temperature using color filters:

10 "10"

ΔM [μrά] = = M ₂ - M.

T ₂ [K] T _χ [K]

With Mi: MIRED value for T _x (in μrd),

M ₂ : MIRED value for T (in μrd),

AM: = M ₂ -M _x : correction value (in μrd),

T _x : given color temperature of the light source (in K),

T ₂ : set color temperature of the camera

(in K) and

ΔT: = T ₂ - T: temperature difference (in K).

The result of such a calculation can be positive or negative. If the result AM is positive (for T> T ₂ ), then a yellow filter (ie a warmer shade) must be selected; a negative result AM (for T _x <T ₂ ) requires a blue filter (ie a colder hue). For example, if the color temperature of the light source is T = 4,706 K and the set color temperature of a camera is T ₂ = 3,200 K (daylight on a clear day just before sunset), the calculation is:

10 ⁶ 10 °

ΔM = - H-100 / -rd

3,200 K 4,706 K

In this case, the filter should have a correction factor of +100, i.e. the reciprocal color temperature of the light source l / T _x , multiplied by 10 ^e , must be reduced by AM = 100 μrd. This means that the set color temperature of the camera T ₂ = 3,200 K must be increased by AT = 1,506 K to the color temperature of the light source T = 4,706 K. Such a filter is also referred to as a WlO filter. The letter in the filter designation indicates whether the color temperature of the camera T _{2 should be} reduced (C) or increased (W). The number after the letter indicates the conversion value in tenths of a μrd. The colors of the video image generated by a video camera should always reflect the original color impression of the recording scene, regardless of the color temperature of the scene lighting. To meet this requirement, the camera must be set so that a white object is always shown in white. This adaptation to the color temperature of the scene lighting is referred to in video technology as a white balance or white balance setting. Basically, video cameras are set so that a white object recorded at a color temperature of T »3,200 K (standard light source) is also imaged as a white object. Whenever the color temperature deviates from this standard value, the white balance of the camera must be changed until the object recorded in white is shown in white again. There are usually three options for white balance adjustment in conventional video cameras:

• manual setting by adjusting potentiometers (in older models), • automatic setting for a given scene lighting and

• the automatic adjustment of the white balance over a wide color temperature range.

Basically, the white balance setting affects the amplification of light in the red and blue signal processing. It becomes clear that this has to be the case if one considers that with a change in color temperature according to higher values (ie with higher color temperature T _{x of} a light source) the blue component, with a change in color temperature after lower values (ie with lower color temperature T _x) a light source) the red component of the light coming in through the lens increases. Most of the video cameras available today have the option of automatic white balance adjustment. It is sufficient to pick up a white object and briefly press an "Auto White" or "Auto White Adjust" button. Pressing this button triggers a control process that changes the gain in the red and blue signal path with the aid of a potentiometer until the red and blue signal voltages U _R -γ and U _B .γ of the color difference signals RY and BY are the same size. The disadvantage of this method for automatic white balance control is that when the scene lighting is changed, a new adjustment with a white object must always be made.

Some newer camera models work with a white balance setting that constantly adapts to the new daytime, weather-dependent or changing ambient light conditions due to artificial light sources. In the terminology of image processing technology, this type of white balance setting is referred to as "auto white tracking" (AWT). With the AWT control, the spectral distribution of the scene lighting is constantly measured by its own sensor, the AWT sensor. The AWT sensor is located close to the lens. The AWT sensor itself consists of individual photodiodes, each of which receives the incident light via a red,

Green and blue filters are fed. If the scene lighting is low, correct readjustment cannot be guaranteed. In this case, a detector detects the underexposure and interrupts the AWT control circuit via an electronic switch. The camera then works automatically with a fixed white balance for a color temperature of T ₂ = 3,200 K. Regarding claim 10:

The transmission-side switchover to the data signal sample value S _{n + k} can take place immediately after a predetermined delay time Δt _{v has} elapsed and / or after at least one suitably defined threshold value σ _re f has been exceeded or fallen short of. This "hard" switchover ensures that the signal receiver reacts quickly to changed signal values. If the capacity C of the transmission channel is large enough and the transmission time Δt _{ü is} short enough, relevant changes As (n + m, m) of the signal s (n) can reach the signal receiver almost in real time.

Regarding claim 11:

The transmission-side switchover to the data signal sample value s _{n + k} can alternatively also gradually after the elapse of a predetermined delay time Δt _v and / or after the over or Falling _below at least one suitably defined threshold value σ _ref . For this purpose, an interpolation method can be used, for example, which brings about a smooth, smooth transition to the new signal value s _{n + k} over several sampling periods. This "gentle" switchover can ensure that the human perception system is not overwhelmed by a sudden sharp increase in the received signal intensity, for example by noise due to a booming loud noise or by glare due to a bright light or is damaged. For example, when loud loud acoustic signals suddenly appear via headphones, permanent hearing damage can easily occur. However, smooth transitions in the received signal intensity can enable the human perception system to slowly get used to the changed signal values, so that there is a risk of damage of human sensory cells can be reduced by a sudden increase in the received signal intensity.

Regarding claim 12: if the duration Δτ of the change in the data signal (DS) falls below a predetermined delay time duration Δt _v , that is, provided

Δτ <Δt _v or K <k at Δt _v ≠ 0 or k ≠ 0 and d = 0 for K: = Aτ / At] and k: = [At _v / At] (with k ≥ 1)

applies, according to the invention there is no regulation of the values of the data signal (DS).

Regarding claim 13:

If the magnitude of the amplitude, frequency or phase si of the data signal (DS) falls below a predetermined amplitude, frequency or phase threshold value at at least one observation time i, that is, provided

sι | | <dl with d: = 2-S _re f / V-? = S _ref ^ 2 at Δt _v = 0

applies, according to the invention there is likewise no regulation of the values of the data signal (DS).

Regarding claims 12 and 13:

If both a delay time period Δt _v (Δt _v ≠ 0) and an amplitude, frequency or phase threshold / 2 (d ≠ 0) are allowed as threshold values, this is done in the cases

a) Δτ <Δtv or K <k for Δt _v ≠ 0 or k ≠ 0 and | sι | <d / 2 for d ≠ 0, b) Δτ <Δt _v or. <k for Δt _v ≠ 0 resp. k ≠ 0 and | -Si | > d / 2 for d ≠ 0 as well

c) Δτ> Δt _v or K> k for Δt _v ≠ 0 or k ≠ 0

according to the invention also no regulation of the values of the data signal (DS). According to the invention, these values are only regulated if the conditions

Δτ> Δt _v or K> k for Δt _v ≠ 0 or k ≠ 0 and | ≤i | > d / 2 for d ≠ 0

be respected.

Regarding claim 14:

The signal component of the data signal (DS) regulated by using this method can be, for example, a chrominance signal (U, V). To simplify the spelling, however, only one data signal (DS) was mentioned in the above formulas, it no longer being distinguished between the data signal (DS) and signal components of this data signal (DS).

Regarding claim 15:

The signal component of the data signal (DS) regulated by the use of this method can be, for example, a luminance signal (Y). To simplify the spelling, however, only one data signal (DS) was mentioned in the above formulas, it no longer being distinguished between the data signal (DS) and signal components of this data signal (DS). Description of figures with list of reference symbols

The invention is described in more detail below on the basis of preferred exemplary embodiments as described in FIGS. 1 to 8.

Show in detail

FIG. 1 shows a block diagram to illustrate the signal transmission from a signal transmitter to a signal receiver in the case of moving picture sequences as a transmission signal,

FIG. 2 shows a flow chart to illustrate the determination, evaluation and transmission of the difference signal Δs (n + -7., - τι) with the delay time Δt _v ,

FIG. 3 shows an example of the formation of a differential signal Δs (n + -7?, - τ?) For a shift by m = 5 sampling times and a delay period of k = 5

sampling times,

4 shows two cases for a persistent signal (that is, s (n) = const.) During the observation period, that is to say the delay time Δt _v between the measurements #i and # (i + k) of the signal s (n) at the sampling times n and n + k,

FIG. 5 shows two cases for a significantly increasing or decreasing signal (ie s (n) ≠ const.) During the observation period, that is to say the delay time Δt _v between the measurements #i and # (i + k) of the signal s ( n) at the sampling times n and n + k, 6 shows a case for a significantly increasing and decreasing signal (ie s (n) ≠ const.) During the observation period, that is to say the delay time Δt _v between the measurements #i and # (i +) of the signal s (n) the sampling times n and n + k,

Figure 7 each a special case for the assumption of a persistent amount of the signal s (n) of the signal duration

Δτ> Δt _v (i.e. K> k) or a non-persistent amount of the signal s (n) of the signal duration Δτ <Δt _v

(ie K <k) without taking into account an amplitude threshold d / 2 (ie d = 0) and

FIG. 8 each a special case for the assumption of a significant amount of the signal s (n) (ie | si |> d / 2) or an insignificant amount of the signal s (n) (ie | Si | <d / 2) ) without considering a delay time Δt _v (i.e. Δt _v = 0 or k = 0).

A block diagram is outlined in FIG. 1, which shows the signal transmission chain from the signal transmitter to a signal receiver, which consists of a transmission branch, a transmission channel and a reception branch. In this case, the signal to be transmitted is a moving image sequence (101) which was recorded by an analog video camera (102). The analog output signal s (t) of the video camera (102) is fed to the input of a module (103) which carries out a brightness control (HR) or a white balance (WA) for the analog transmission signal s (t) performs. This can ensure that brief spikes or dips and / or insignificant signal changes in the transmission signal s (t) are not transmitted. The module (103) can be integrated, for example, as a microchip in the video camera (102). After scanning and Quantization of the analog transmit signal s (t) by the analog-digital converter (104) produces the digital transmit signal at its output

^« -, [l for n = θ] _^ s (n) = Ys (m) - δ (n- m) for δ (n): = \} V ne IN ₀ , _o {0 otherwise J dh s (n ) e {0, ..., 0, s ₀ , Si, ..., s _n , ..., s _N , 0, ..., θ}.

If a digital video camera is used instead of an analog video camera, component 104 can be dispensed with. This signal is digitally encoded for the purpose of reducing the redundancy in the encoder (105), a difference signal Δs (n + m, m) being determined, the values of which are subtracted from the digital transmission signal s in) by one by one sampling time (-71 = 1 ) shifted digital transmission signal s (n + m) results in:

In addition to the source coding, channel coding can also be carried out here. If the signal Δs (j - + - 7i, .τ?) Is to be transmitted via the air interface, it must first be modulated onto an analog high-frequency carrier signal with the aid of a modulator (106). The transmission channel is analog in this case and can be modeled by adding (107) a white Gaussian noise signal r (t)

(AWGN channel). To simplify the formulas, it should be assumed below that the transmission channel is ideal, that is, r (t) = 0. In this case, the difference signal As (n + m, m) at the output of the demodulator (108) can be completely in the signal receiver and be reconstructed without errors. If the difference signal Δs (n + -? Ι, -77) is not via the air interface, but Transfer wired, components 106, 107 and 108 in the outlined signal transmission chain are omitted. The transmission channel, ie the conductive connection between the transmitter-side encoder (105) and the receiver-side decoder (109), is then digital. After decoding (109) and digital-analog conversion (110), the moving image sequence can be displayed on a screen or a display (111) and perceived by the eye of an observer (112). If a digital screen or a digital display is used instead of an analog screen or an analog display, then component 110 can be dispensed with.

In addition to the sketched version of the signal transmission chain, a version is also conceivable in which the module (103) for performing the brightness regulation (HR) or for performing the white balance (WA) for the analog transmission signal s (t) behind the output of the encoder (105) is arranged. In this version, the brightness regulation or the white balance takes place after the determination of the difference signal Δs (n + -τ., Ii.) Has been completed, that is to say for signals in digital form. The figures in the following figures 2 to 8 relate to this version.

FIG. 2 shows the flow of the determination, evaluation and transmission of the difference signal As (n + m, m) with the delay time Δt _v according to the invention in a flow chart. Before the beginning of the procedure shown here, a digital signal s (n) is present, which is obtained by sampling with the sampling period Δt while observing the Shannon theorem

Δt <l / (2-B),

and quantization of an analog useful signal s (t) of bandwidth B was obtained. With this notation, the Number n of the respective sampling time t _n = n - At is abbreviated for this time. For drawing technology reasons, however, only one subsampled signal s in) is shown in FIG. 2, that is to say a signal for which the Shannon theorem is not fulfilled.

After the initialization of the counter Z _A for the sampling times n by the assignment n: = 0 and the initialization of the counter Z _M for the measurement times i with the aid of the assignment i: = 0 (step 201), a signal is carried out on the signal transmitter side first measurement (measurement #i) and storage of the current signal value of s {n) at the sampling time n (step 202). This measured sample value of the digital signal s (n) is referred to below as si: = s _n . If a delay time .DELTA.t _{v has been} waited for (step 203), which may comprise a few sampling periods .DELTA.t (.DELTA.t _v > .DELTA.t), the counter Z _A is incremented for the sampling times n by executing the assignment n: = n + k with the shift

k: = [At _v / At} ≥ 1,

an incrementation of the counter Z _M for the measuring times i by performing the assignment i: = i + k (step 204) and a measurement and storage of the current signal value of s {n + k) at the sampling time n + k (step 205) in the signal transmitter made (measurement # (i + k)). During the delay time Δt _v , further measurements and storage of sampled signal values sι ₊ ι: = s _{n + 1} , Si ₊₂ : = s _{n + 2} to Si ₊ -ι: = s _{π + k} -ι (measurements # (i + l) to # (i + -tl)). The sampled signal value at the time of the (i + k) th measurement is referred to below as Si _{+ k} : = s _{n + k} .

The difference signal Δs (n + m, m) to be transmitted is generally regarded as the signed difference of the m sampling times delayed useful signal s {n + m) compared to the original signal s in) measured in the signal transmitter according to the regulation

s (n + m) -s (n) for O ≤ n ≤ N - m

0 otherwise

calculated. The signal change Δs _{+ k} to be transmitted at the time n + _k is accordingly according to the formula

k

Δs _{i + k} : = Si ₊ k - Si = ^ Δs. ^

3 = 1

determined (step 106). Next, a query (step 107) is used to test whether there are already two samples of the signal s (n) at a distance m that can be compared with one another. If at least two samples of the original signal s {n) were waited for, the following applies

n + m ≥ 2 ^• m <-> n ≥ m,

a comparison of two samples Si _{+ j} and si _{+ j} -i (for j = 1,2,3, ..., k) of the signal s (n) at times n + 1, n + 2, n + 3 to n + k to form a difference signal value

S _{+ j} : = Si _{+ j} - i _{+ j} -i

be made ("Yes" case). Otherwise, a return is made to step 103 ("No" case). In the "yes" case, it is necessary to examine whether the signal values Si, si + i, s _{i + 2} determined at the sampling times n, n + 1, n + 2, n + 3 to n + k to s _{i + k} a predetermined threshold value + S _ref / ^' V2 or

exceed or fall below, i.e. the inequality | s (n) | > 5 _ref / V-? for n, n + 1, n + 2,. , , , n + k (*) alSO | Si | > Sre _f / V-? Λ | Si ₊ χ | > S _re f / 2 Λ. , , Λ | Si ₊ k | > 5 _re f / V-?

is satisfied . The amount of the distance between the upper amplitude threshold + S _ref / V2 and the lower amplitude threshold -S _r e _f / -? be with in the following

: = 2 • S _ref / 2 = Sref ₂

designated. A query (step 108) is used to test whether the above inequalities (*) apply or not. If they apply ("yes" case), the difference signal value Δsi ₊ can be transmitted from the signal transmitter to the signal receiver at the sampling time n + k. Otherwise, a return is made to step 103 ("No" case). On the receiver side, after the transmission time .DELTA.t _{ü has} elapsed, the sample value of the difference signal .DELTA.si _{+ obtained with the} correct sign is added to the last available sample value Si of the signal s in) in order to obtain the signal value at the sampling time n + k

Si ₊ k = Si + Δsi _{+ k}

to obtain. If the transmission of the complete difference signal Δs (ι - + - τ., - τ.) Has ended, ie the inequality

n> N-m

the transfer is complete. Otherwise, the procedure starts again from step 101.

In FIG. 3, an example consisting of three figures shows how the difference signals Δs (n + -7?, - τ.) be communicated. In Figure 1, the analog signal voltage s {t) is considered as the output signal, the instantaneous amplitude of which, for example, indicates a value proportional to the instantaneous brightness or volume of a recorded image or sound signal. After the sampling of the analog useful signal s {t) with the sampling period Δt (normally in compliance with the Shannon theorem, but here for reasons of drawing technology in a subsampled representation) and the quantization of the sampled signal, the useful signal is available in digital form s {n) as shown in Figure 2. In order to reduce the data rate R of the signal transmission from the signal transmitter to the signal receiver, not the signal s {n), but the difference signal As {n + m, m) delayed by m sampling times) is transmitted. Figure 3 shows the difference signal As {n + m, m) determined for the shift m = 5. The delay in the transmission of Δs (n + -τ., - τ.) Is also necessary in order to be able to neglect briefly occurring interference peaks in the original signal s (n) which may occur within the k sampling times concerned. The signal s (n) is quasi-stationary during this period, so it applies

ds (t)

«0 for t e [n - At; (n + k) - At], German

the difference signal value of is transmitted

As (n + m, m) at time n + k only if s (n) at that time has a suitably determined amount

Threshold value S _ref / xj ^~ 2 exceeds. This threshold value can be determined by a direct voltage signal s _ref (t) = const. or represented by the effective value of the amplitude of a sinusoidal AC voltage signal Sref (t), which serves as a reference signal. FIG. 4 illustrates two cases for a persistent signal (s (n) = const.) During the observation period, that is, the delay time Δt _v between the measurements #i and # (i + k) of the signal s (n) at the sampling times. ten n = 15 to n + k = 20.

In case a) an example is shown in which the magnitude of the signal amplitude of s (n) assumes a constant value above a predetermined threshold value for said (k + 1) sampling times. For the decision as to whether the current sample value Si _{+ k of} the signal s (n) should be transmitted at the sampling time n + k = 20, only the determined values of the signal s (n) at the time of the measurements #i to # (i + k), i.e. at the sampling times n = 15 to n + k = 20. Because in case a) the condition

at times n = 15 to n + k = 20 (* *)

is satisfied, the transmission of the difference signal value Δsi ₊ k can take place at the time n + k = 20.

In case b) an example is shown in which the magnitude of the signal amplitude of s (n) for the said {k + 1) sampling times assumes a constant value below a predetermined threshold value, namely the value zero. Only the determined values of the signal s (n) at the time of the measurements #i to # (i + k) are of interest for the decision as to whether the current sample value Δsi _{+ k of} the difference signal should be transmitted at the sampling time n + k = 20. , i.e. at the sampling times n = 15 to n + k = 20. Since the condition (**) is not fulfilled in case b), the differential signal value Δs _{i + k} cannot be transmitted at the time n + k = 20. FIG. 5 shows two cases for a significantly increasing or decreasing signal (s (n) ≠ const.) During the observation period, that is to say the delay time Δt _v between the measurements #i and # (i + k) of the signal s in) Sampling times n = 15 to n + k = 20. The term "significant" in this context means that the magnitude of the signal amplitude of s (n) within the said (k + 1) sampling times exceeds or falls below a predetermined threshold value. In case c) there is a significant increase in the absolute signal change | s (n) | represented during this period, in case d) a significant decrease of \ s (n) \. During this period, the signal s (n) assumes values for which

s (n)> + S _{X Ϊ} / SI ^~ 2 resp. 0 <s (n) <+ S _ref / 2

applies, whereby there is an overshoot or undershoot of the

Threshold +5 _ref / v2 is coming. However, cases are also possible in which the signal s (n) assumes values during this period for which

s (n) <- S _ret / 2 resp. 0> s (n)> - 5 _ref 'V-?

applies, whereby the threshold value is exceeded or not reached

comes. For the decision as to whether the difference signal value Δsι ₊ should be transmitted at the sampling time ι- + A: = 20, only the determined values of the signal s (n) at the time of the measurements #i to # (i + k) are of interest at the sampling times n- = 15 to n + k = 20. Since the condition (**) is not fulfilled in case c), the differential signal value Δs _{i + k} cannot be transmitted at the time n + k. In case d), the condition (**) is also not fulfilled, so that here too the transmission of the difference signal value Δs _{i + k} cannot take place at the time n + k = 20. The case e) shown in FIG. 6 shows an example of a significantly decreasing and again significantly increasing signal (s (n) ≠ const.) During the observation period, that is to say the delay time Δt _v between the measurements #i and # (i + k) of the signal s (n) at the sampling times n = 15 and n + k = 20. For the decision as to whether the difference signal value Δsi _{+ k} should be transmitted at the sampling time n + k = 2 Q, again only the determined values of the signal s (n) at the time of the measurements #i to # (i + k) are of interest the sampling times n = 15 to n + k = 20. Since the condition (**) is fulfilled in case e), the difference signal value Δsi _{+ k} can be transmitted at time n + k, regardless of which values s {n) for the sampling times n + l = 16 to n in between + kl = 19 assumes.

In Figure 7, two special cases are outlined, in which only the delay time Δt _v , but no amplitude value d / 2 is permitted as a threshold value; so d = 0 applies. If the signal s (n) for the signal duration Δτ or K: = \ ^~ Aτ / At] with

Δτ> Δt _v or with K> k at d = 0 (assumption 1)

is persistent, the difference signal value Δs + is transmitted at the sampling time n + k = 20. However, this applies to the signal duration

Δτ <Δt _v or K <k at d = 0 (assumption 2),

the differential signal value Δs _{i + k} is not transmitted at the sampling time n + k = 20. In Figure 8, two special cases are outlined, in which only the amplitude value d / 2, but no delay time Δt _{v is allowed} as a threshold value; therefore Δt _v = 0. If the magnitude of a signal value Si exceeds an amplitude value d / 2, that is

| Si | > d / 2 at Δt _v = 0 (assumption 3)

applies, the difference signal value Δsi _{+ k} is transmitted at the sampling time n + k = 20. However, this applies to the signal duration

) Si | <d / 2 at Δt _v = 0 (assumption 4),

the differential signal value Δs _{+ k} is not transmitted at the sampling time n + k = 20.

The meaning of the symbols denoted by numerals in FIGS. 1 and 2 can be found in the list of reference symbols below.

LIST OF REFERENCE NUMBERS

No.

101 block of houses with sidewalk as recorded image scene of a moving image sequence

102 Video camera as a signal recorder for recording moving picture sequences with the analog output signal s (t) as a transmission signal

103 Module for performing a brightness control (HR) or for performing a white balance (WA) for the analog transmission signal s {t)

104 analog-digital converter for sampling and quantizing the analog transmission signal s {t) with the digital transmission signal s {n) at the output

105 encoders for performing the source and / or channel coding of the digital transmission signal s (n) with the difference signal fs (n + m) - s (n) for 0 <n ≤ N - m As {n + m,): = < with m = 1 0 otherwise at the encoder output

106 modulator for carrying the differential signal As (n + m, m) to be transmitted with the output signal Δs _mθ d (t)

107 adders for modeling the disturbance of the signal Δs _mod (t) to be transmitted due to an additive white Gaussian noise r (t)

(To simplify the formulas, the assumption below is that the transmission channel is ideal, i.e. r (t) = 0.)

108 demodulator for recovering the differential signal Δs to be transmitted (2- + iT., - 7.)

109 decoders for performing the source and / or channel decoding to obtain the digital transmission signal s (21 + 27.) = As (n + m, m) + s (n) at the output of the decoder

110 digital-to-analog converter for the reconstruction of the analog transmission signal s (t)

111 Screen or display as signal receiver for displaying the recorded moving picture sequences

112 Perception of the received moving picture sequences in the eye of the beholder

201 Actions: Initialization of the counter Z _A for the sampling times 2 by the assignment n: = 0 and initialization of the counter Z _M for the measurement times i by the assignment i: = 0

202 Actions: Measurement of the sampled signal value Si: = s _{n of} a digital useful signal s (n) at the sampling time n in the signal transmitter (measurement #i) and storage of this value 203 Actions: Wait for a delay time Δt _v (Δt _v ≥ Δt) comprising several sampling times while performing further measurements of sampled signal values s _{i + 1} : = s _{n + 1} to si _{+ k} -i: = s _{n + k} -ι (measurements # (i + l) to # (i + Jc-l)) and storing these values

204 Actions: incrementing the counter Z _R for the sampling times 21 by the assignment 21: = n + k with the shift k: = rΔtv / Δtl> 1 and incrementing the counter Z _M for the measurement times i while carrying out the assignment i: = i + k

205 Actions: Measurement of the sampled signal value Si _{+ k} : = s _{n + k of} the useful signal s (n) at the sampling time n + k in the signal transmitter (measurement # (! +)) And storage of this value

206 Action: Determination of the current signal change to be transmitted

Δsi _{+ k} : = S _{+ k} - Si of the signal value Si ₊ from s (n) at the sampling time n + k compared to the signal value S from s (n) at the sampling time n in the signal transmitter 207 Query: Have at least two samples of the original signal s (n) waited, does n + m ≥ 2 - m <=> n ≥ m apply?

208 Query: The signal values Si, si ₊ i, Si ₊ , ..., Si _{+ k} determined at the sampling times n, 21 + I, 21 + 2, ..., n + k fall below a given threshold value + S _ref / V-? or - S _τe t / 2, the inequality | si ₊ j | > S _ref / * J ^~ , that is | s _{i + j} | > d / 2 for j = 0,1,2,. , , k 1

²⁰⁹ Action: Transmission of the current signal change Δsi _{+ k} from the signal transmitter to the signal receiver at the sampling time n + k and implementation of the addition

210 Query: Has the transmission of the complete difference signal Δs (n + m, m) ended, does n _> Nm apply?

Claims

claims

1. A method for regulating at least one signal component of at least one data signal (DS), characterized in that the regulation is carried out by means of at least one correction signal (KS) at discrete times, depending on the amplitude or duration of the changes in at least one signal component of the data signal ( DS).

2. The method according to claim 1, characterized in that the data signal (DS) is transmitted from a signal source (102) to a signal sink (111) by means of a transmission of differential data signals (DDS).

3. The method according to claim 1 or 2, characterized in that the data signal (DS) changes its current value in the course of the regulation, provided that the values of the data signal (DS) exceed or exceed at least one threshold value σ _ref suitable as a reaction level. below.

4. The method according to claim 3, characterized in that at least one of the threshold values σ _ref used as the reaction level is a predefined delay time period Δt _v .

5. The method according to any one of claims 3 or 4, characterized in that at least one of the threshold values σ _ref used as the reaction level is an amplitude, frequency and / or phase value.

6. The method according to any one of claims 1 to 5, characterized in that the data signal (DS) is a video signal (VS).

7. The method according to any one of claims 3 to 6, characterized in that the intensity and / or frequency given as the threshold value is at least one luminance (Y _ref ) and / or chrominance value (U- _{ref /} V _ref ) with which the values of the luminance (Y) or chrominance (U, V) of a received digital video signal (VS) are compared.

8. The method according to any one of claims 3 to 7, characterized in that to determine the threshold values for the smallest perceptible luminance (ΔY _J N) or

Chrominance change (ΔL7 _JN , ΔIJN) of the received video signal (VS) as a function of the lighting conditions in the environment of the image signal generator characterized by (Y, U, V) _env , the psycho-optical properties of human eyesight with regard to the amplitude and / or frequency modulation threshold , the perceptible frequency group width Δf _G and the effects caused by the masking of useful radiation by interference radiation in the time and / or frequency range are evaluated.

9. The method according to any one of claims 3 to 8, characterized in that it is to carry out the white balance of video signals (VS), in the brightness, hue and / or color saturation of the moving images recorded by a video camera (102) (101) can be regulated depending on the lighting conditions in the vicinity of the video camera (102).

10.The method according to any one of claims 3 to 9, characterized in that after the lapse of a delay time .DELTA.t _v and / or after exceeding or falling below at least one suitably defined threshold value σ _ref, an immediate switch to the new value of Correction signal (KS) takes place.

11. The method according to any one of claims 3 to 9, characterized in that after the lapse of a delay time .DELTA.t _v and / or exceeding or falling below at least one suitably defined threshold value σ _re f by means of an interpolation method, a gradual switchover to the new value on the transmission side of the correction signal (KS).

12. The method according to claim 1, characterized in that a control does not take place if the duration of the change in the data signal (DS) falls below a predetermined delay period (Δt _v ).

13.The method according to any one of claims 1 or 12, characterized in that a control does not take place if the amount of the amplitude, frequency or phase of the data signal (DS) falls below a predetermined amplitude, frequency or phase threshold.

14.The method according to any one of claims 12 or 13, wherein the signal component is a chrominance signal.

15. The method according to any one of claims 12 to 14, wherein the signal component is a luminance signal.