WO2005069600A1

WO2005069600A1 - Adjustment device and method for the colour correction of digital image data

Info

Publication number: WO2005069600A1
Application number: PCT/EP2004/013216
Authority: WO
Inventors: Klaus Anderle
Original assignee: Thomson Licensing
Priority date: 2004-01-08
Filing date: 2004-11-22
Publication date: 2005-07-28
Also published as: DE102004001295A1

Abstract

An adjustment device for the correction of colour values of digital image data has a computing unit, which, in terms of program technology, is set up for calculating secondary colour values from primary colour values. The primary colour values are related to the first representation means and the secondary colour values are related to a second representation means. The adjustment device comprises a memory containing a table in which each primary colour value is assigned a secondary colour value. An interpolation stage is provided, which calculates by interpolation secondary colour values for such primary colour values which are not contained in the table. Furthermore, a method for calculating an assignment table between primary and secondary colour values is proposed.

Description

Adjustment device and method for the colour correction of digital image data

The invention relates to an adjustment device and to a method for the colour correction of digital image data.

What is of crucial importance to the viewer of a film or television production, besides picture sharpness, is the subjective colour impression. Therefore, in the course of producing the productions, it is endeavoured to ensure that the colours, when reproduced at the cinema and/or on a television screen, as far as possible appear in the way that the director intended. This aim, inter alia, is striven for with the postprocessing of film or image material. A prerequisite for an efficient postprocessing is that, by way of example, the colour representation on the monitors of a colourist corresponds as exactly as possible to the image projected in a cinema, for example. Nowadays, the starting point for postprocessing is generally digitized image data generated by film scanners or electronic cameras. Added to these are computer-generated images which are present as digital image data from the outset.

Devices which strive for such correspondence between colour representations with different representation means are already commercially available as software and hardware solutions. These devices are based ^' on the considerations described below.

Colours arise in different ways on different reproduction media. From the earliest times it has been known from painting that from just three different pigments, namely yellow, blue-green and purple-red, all intermediate hues can be produced by mixing the primary colours mentioned. Primary colours are understood to be those colours which cannot be mixed from other colours but from which all other colours can be mixed. In chromatics nowadays this type of colour mixing is referred to as subtractive colour mixing. The term subtractive colour mixing is derived from the fact that a pigment layer absorbs certain spectral components of incident white light and reflects others, as a result of which the colour impression arises for the viewer. Other types of colour mixing were initially not known.

It was not until a long time later that Isaac Newton recognized that the spectral colours of light, the so-called colour stimuli, can also be mixed. With this type of colour mixing the jargon uses the term additive mixing in contrast to subtractive colour mixing explained above in the case of pigments. Additive colour mixing is governed by relatively simple rules, known as Grassmann's laws, which also apply to self- luminous screens, such as, for example, monitors based on cathode ray tubes.

A special case of subtractive colour mixing is the combination or superposition of optical filters. The transmission of the filter combination is equal to the product of the respective transmissions of the individual filters, which is why the jargon also uses the term multiplicative colour mixing in this case. This last-mentioned type of colour mixing is also critical for colour reproduction in the projection of colour films which have three different colour layers lying one above the other.

Figure 1 diagrammatically shows an example of the construction of a colour film 1 in cross section. A layer carrier 2 carries three colour layers 3, 4, 5 having the primary colours red, green and blue, the red-sensitive colour layer 3 adjoining the layer carrier 2 and the blue-sensitive colour layer 5 forming the topmost colour layer. A yellow filter 6 lies between the blue-sensitive and green-sensitive colour layers 5 and 4, respectively. For the purpose of better illustration, the individual layers are represented spaced apart in figure 1 but in reality they adjoin one another. The intermediate layer for preventing interdiffusion of the green-sensitive and red-sensitive colourants is not taken into account here and is not illustrated in figure 1 since it has no influence on the colour behavior of the film which is essential to the present invention.

One important difference between additive and multiplicative colour mixing is that Grassmann' s laws cannot be applied to multiplicative colour mixing. The reason for this is to be found in the fact that, by way of example, as the thickness of a cyan filter increases, there is a decrease not only in the transmission in the red spectral range but also to a considerable extent in the green spectral range. This fact and the resulting consequences are explained in detail further below. In known colour correction systems, therefore, the absorption of test patterns ("test patches") is measured with the aid of densitometers and the absorption in the secondary densities is corrected by a transformation of the colour coordinates.

The colour coordinates are transformed for example by means of so-called look-up tables in which an input colour value is assigned an output colour value, thereby obtaining a conversion or transformation of the colour values. The more extensive the table, the greater the resolution of the conversion in the colour space. In practice this means that the colour conversion is all the more precise, the more extensive the table. The content of the table depends on the parameters in accordance with which the colour conversion is intended to be carried out. These parameters are prescribed by the colourist or altered during the processing of the film material. A change in the parameters has the consequence that the output values of the table have to be recalculated on the basis of the new set of parameters. A given computing power consequently limits the size of the table that can be made available as it were continuously to a colourist. Only in this case does the colourist see the effect of a parameter change without an appreciable delay on his screen, which facilitates his work.

Taking this as a departure point, it is an object of the invention to provide an adjustment device with which good results are obtained during the colour correction even given a limited resolution of a colour space.

This object is achieved by means of an adjustment device according to Claim 1.

The adjustment device according to the invention for the correction of colour values of digital image data has a computing unit, which, in terms of program technology, is set up for calculating secondary colour values from primary colour values.

The primary colour values are related to the first representation means and the secondary colour values are related to a second representation means. Moreover, the adjustment device comprises a memory containing a table in which each primary colour value is assigned a secondary colour value. Furthermore, an interpolation stage (36) is provided, which calculates by interpolation secondary colour values for such primary colour values which are not contained in the table.

In one exemplary embodiment of the adjustment device, the primary colour values may be divided into a first and a second group of colour values. In this case, the first group of primary colour values preferably describes such points in the colour space for which a secondary colour value is stored in the table.

Moreover, it is expedient if the second group of primary colour values describes such points in the colour space for which no secondary colour value is stored in the table.

It is a further object of the invention to specify a method for calculating an assignment table of primary colour values.

This object is achieved by means of a method in accordance with the independent method claim.

The method according to the invention for the colour correction of digital image data assigned to a first representation means comprises the following steps:

(a) calculation of a table with a first number of primary colour values that are in each case assigned a secondary colour value assigned to a second representation means; (b) storage of the table in a memory;

(c) separation of incoming primary colour values into those for which a secondary colour value is stored in the table and those for which no secondary colour value is stored in the table; (d) calculation of further secondary colour values in order to increase the number of primary colour values in the table that are in each case assigned a secondary colour value; and (e) repetition of steps (b) , (c) and (d) until the largest possible number of primary and secondary colour values are stored in the table, which can be achieved with a predetermined length of the data values for the colour values. In one embodiment of the method according to the invention, it begins anew in the event of a change in the parameters on which the calculation is based.

It may also be advantageous if the calculation of further secondary colour values depends on the available computing power.

The drawing illustrates facts which serve to provide a better understanding of the invention, and also an exemplary embodiment of the adjustment device according to the invention. In the Figures:

Figure 1 diagrammatically shows the structure of a colour film in cross section,

Figure 2 shows the construction of a colourist' s workstation in greatly simplified form,

Figure 3 shows the spectral density of the blue, green and red colour layers of a colour film,

Figure 4 shows a flow chart of the method according to the invention, and

Figure 5 shows an adjustment device according to the invention for the correction of colour values.

Figure 2 illustrates a colourist' s workstation in greatly simplified form. In the course of film production, a first copy is made from the film material originally exposed by the camera. The copy is used to produce further prints which form the starting point for the postprocessing of the film. In Figure 2, such a print is inserted in a film scanner 11. During the scanning of the print, the photographic image information is converted into digital image data and fed to a device 12 for colour correction, which is usually operated by a colourist. During the correction of the film material, the colourist views the image to be processed on a monitor 13. The colour representation on the monitor 13 is determined by colour values at the output of the colour correction device. The colour values at the output of the colour correction device 12 are also forwarded as control commands or "Code Values" to a film exposer 14, which exposes the data onto an internegative film. The content of the internegative film is then transferred to a positive film by means of a contact copy. The positive film is symbolized by a film reel 16 in Figure 2. In order to inspect the result of the exposed film, the latter is projected onto a projection screen 18 by a film projector 17. Ideally, the colour representation of an image projected onto the projection screen 18 corresponds to the colour representation of the same image on the monitor 13. For approximation to this ideal case, a device 19 for adjusting the colour coordinates is connected between the colour correction device 12 and the monitor 13. The adjustment device 19 converts the "Code Values" sent to the film exposer 14 into colour coordinates for the monitor 13. The conversion has the aim of obtaining as far as possible identical colour representations on the monitor 13 and the projection screen 18, respectively. The conversion method and the conversion device 19 are described in greater detail below.

Figure 3 illustrates spectral curves of in each case three colour filters of different density for the colours red, green and blue. The density D is plotted on the ordinate and the wavelengths in nanometers (nm) are plotted on the abscissa. The density D of a filter is derived from the transmission T thereof in accordance with the following formula: D = -log (T)

This means that at density zero, the relevant filter is completely transparent, and that the transmission decreases as the density increases. Density curves for filters with different transmissions are plotted for each of the primary colours red, green and blue. It can clearly be seen that, for the density curves for the red filter, by way of example, appreciable secondary maxima occur in the blue spectral range around 400 n , and lead to a considerable absorption for the colour impression. The same applies to a lesser extent to the density curves of the green filters. The density curves for the blue filters fall sharply in the wavelength range of between 440 nm and 380 nm only to rise again below 380 nm. Furthermore, the density curves of the blue filters, with increasing density, exhibit a more and more highly pronounced plateau in the green spectral range around 550 nanometers, the plateau projecting right into the red spectral range. The absorption of a primary colour filter in spectral ranges other than the spectral range assigned to the respective primary colour is referred to as the "secondary density" of the density curve and results in colour shifts during the projection of colour films for example in the case of multiplicative colour mixing. These effects are known in principle and are corrected for example by means of a linear transformation of the colour coordinates. In order to better understand the extent to which the invention goes beyond the known methods, it is necessary firstly to discuss the conventional correction method in more detail.

Different film materials differ inter alia in the absorption properties of the colourants, which makes it necessary to adjust the colour correction device 12 shown in Figure 2 to a specific film material. For this purpose, the film exposer 14 exposes with predetermined code values so-called "test patches" i.e. image windows with different colours and colour densities. This film material is then copied and produces the actual film. The test patches are then measured by densitometers in order to determine the absorption of a colourant in specific wavelength windows. The measurement characteristic of the densitometers is determined in accordance with DIN 4512 3 or a corresponding international standard. With densitometer measurements, the absorption of the colourants is ascertained not only in the primary maxima but also in the secondary maxima. The values determined in this way form the basis for the subsequent transformation of the colour values which define the representation on the colourist' s monitor 13. The transformed colour values are corrected colour values which define the illumination commands of the film exposer 14 and thus determine the subsequent colour representation on the projection screen 18. To put it another way, the colour values or code values which control the film exposer 14 are "predistorted" in order to compensate for the "distorting" influence of the colourants of the film material used.

It has been shown in practice, however, that the correspondence between the colour representation on the monitor 13 and the projection screen 18 that is striven for in this way still leaves something to be desired. The purpose of the invention is to improve said correspondence.

In order to realize this aim, the invention commences at determination of the correction values. From the more precise consideration of the spectral density curves of the colour filters as shown in Figure 3, it is possible to derive further properties of the colourants which lead to colour shifts. However, these properties cannot be identified by means of the densitometer measurements used in practice. This is because conventional densitometers permit only an integral consideration of the absorption properties of the colourants. Upon more precise consideration of the spectral absorption curves, a shift in the primary maxima toward shorter wavelengths can be discerned for all primary colours as the density increases. This shift S is represented using the example of the primary maximum for red in Figure 3. Furthermore, the form of the density curves also changes as a function of the densities. It is exactly in this way that it is thus possible to determine and correspondingly describe the spectral influences of particular film treatments during the copying process and the development.

In the case of conventional densitometer measurements, these changes are registered only as a change in the absorption in the respective measurement window. For this reason, it is not possible with densitometer measurements to determine the actual absorption at a specific wavelength. However, this is exactly what is important for as precise a correspondence as possible between the colour representation on different representation means.

The invention therefore proposes measuring the test patches of the film materials using a spectrometer over the entire wavelength range and interpolating intermediate spectra from the spectra thus obtained. From the totality of the spectra, it is possible to derive, for the three primary colours, tables which put a colour value that determines the representation on the colourist' s monitor 13 into a relationship with a code value of the film exposer 14. A three-dimensional table is produced overall in this way.

The method according to the invention is described in greater detail below with reference to figure 4. The starting point is formed by RGB colour values which are output from the colour correction device 12 to the monitor 13, on the one hand, and to the film exposer 14, on the other hand. In order to obtain a standardized colour reproduction on the monitor 13, a so-called look-up table for the monitor LUT (M) is stored in the adjustment device 19, said table taking account of the reproduction properties of the monitor. In accordance with the flowchart in Figure 4, the film is exposed in the film exposer in accordance with these RGB values. Said film is then copied onto the material to be projected. The colour patterns or patches generated in this way are measured spectrally in a step 22. In addition to these measured spectra, further intermediate spectra are calculated in a step 23. The totality of the spectra generated in this way are convolved with the perception curves of a standard observer in a step 26 in order to generate colour coordinates X, Y, Z corresponding to the RGB values. The colour coordinates X, Y, Z are finally linked with an "inverted" look-up table of the monitor LUT (M) ^~1 in a step 27. This produces new colour values R' , G' , B' . The influence of the film material on the colour reproduction can be derived from the differences between the colour values R, G, B and R' , G' , B' . Further look-up tables are therefore generated from said differences and are stored in the adjustment device 19 and kept ready for application to the colour values RGB. What is achieved in this way is that the colour representation on the monitor 13 corresponds very well to the colour representation on the projection screen 18.

For generating the three-dimensional look-up table 19, the volume of data increases correspondingly with the third power of the resolution; the number of calculations for generating these data rises with the third power of the resolution. For the final colour adjustment, although it is necessary for the number of support points in the three-dimensional table to be as large as possible, a large number of support points prevents a dynamic colour correction during the film processing. With the computers that are available at the present time, a dynamic correction is not possible owing to the high number of required computation operations. In this case, "dynamic correction" means that the 3D look-up table is adapted to changed parameters input by the colourist at the colour correction device.

Figure 5 diagrammatically illustrates an exemplary embodiment of the adjustment device 19 according to the invention for colour correction, which is shown only as a single block 19 in Figure 2. A three-dimensional look-up table is calculated in a computer 31 and stored in a memory 32. The resolution with which the computer was able to calculate the three-dimensional look-up table defines the number of "most significant bits" MSB. The MSB describe the resolution in the colour space for which calculated support points are present in the table. All intermediate values in the table are determined by interpolation between the support points. All customary interpolation methods which are known in the prior art and prove to be suitable for the present purpose are taken into consideration for this.

In accordance with the boundary between MSB and LSB that is defined in this way, the colour values Rι_n, Gj__n, Bi_n of incoming image data Dj_.n are divided into two data streams by means of a data processing stage 33.

A first data stream contains colour values R_MSBΛ G_MSB, B_MSB in the case of which all LSB are set to be equal to zero. For each colour value R_MSB, G_MSB, B_MSB there is a complementary colour value R_LSB, G_LSB and B_LSB, respectively, which contains the LSB of the complementary colour value and whose MSB are set to be equal to zero. The information originally contained in each individual colour value R_in, Gι_n and Bι_n is thus divided between two colour values, namely R_MSB and R_LSBΛ G_MSB and G_LSB_Λ and B_MSB and B_LSB, respectively. All the colour values mentioned have the same bit length.

In this case, the colour values R_MSBΛ G_MSB and B_MSB correspond to those points in the colour space for which support points have been calculated by the computer 31. The complementary colour values R_LSB, G_LSB and B_LSB correspond to points in the colour space that lie between the said support points.

In a second data processing stage 34, the colour values R_MSBA G_MSB and B_MSB are converted into corrected colour values R _SB, G_MSB and B_MSB in accordance with the support points in the look-up table. In a subsequent interpolation stage 36, the intermediate values corresponding to the colour values R_SBA G_LSB and B_LSB are interpolated between the support point values, so that corrected output colour values RO_ut_r G'_out, B'_out for all the input colour values Rj_.n, Gι_n, Bι_n are available at the output of the interpolation stage 36.

The resolution of the look-up table is increased to the extent to which time for calculating further support points is available to the computer, so that the boundary of the MSB shifts further until the full resolution is ultimately attained in the colour space. If the colourist alters an input parameter of the colour correction, the calculation is repeated anew.

One advantage of the device according to the invention is that the colourist sees the result of a parameter change immediately on the screen 13, albeit only with a reduced resolution in the colour space which, however, generally suffices for the first editing cuts of the film material. The immediate visibility facilitates the colourist' s work.

Claims

Patent Claims

1. Adjustment device for the correction of colour values of digital image data having a computing unit, which, in terms of program technology, is set up for calculating secondary colour values from primary colour values, the primary colour values (Rχ_n, G±_{n r} B_in) being related to the first representation means and the secondary colour values (R'out, G'out, B'out) being related to a second representation means, having a memory containing a table in which each primary colour value (Ri_n, Gι_n, Bι_n) is assigned a secondary colour value (R'out, G'out, B'out) having an interpolation stage (36) , which calculates by interpolation secondary colour values for such primary colour values {R±_n, G_in, B_in) which are not contained in the table.

2. Adjustment device according to Patent Claim 1, characterized in that the primary colour values are divided into a first and a second group of colour values RMSBJ GMSB/ B_MS_B; RLSBC G_LSBΛ B SB) •

3. Adjustment device according to Patent Claim 1, characterized in that the first group of primary colour values (R SBJ G_MSB^ B_MSB) describes such points in the colour space for which a secondary colour value (R'_MSB^ G' _SBΛ B' _SE) is stored in the table.

4. Adjustment device according to Patent Claim 1, characterized in that the second group of primary colour values (RL_SB, G_LSB and B_LSB) describes such points in the colour space for which no secondary colour value is stored in the table.

5. Method for the colour correction of digital image data assigned to a first representation means, the method comprising the following steps: (a) calculation of a table with a first number of primary colour values that are in each case assigned a secondary colour value assigned to a second representation means; (b) storage of the table in a memory; (c) separation of incoming primary colour values into those for which a secondary colour value is stored in the table and those for which no secondary colour value is stored in the table; (d) calculation of further secondary colour values in order to increase the number of primary colour values in the table that are in each case assigned a secondary colour value; and (e) repetition of steps (b) , (c) and (d) until the largest possible number of primary and secondary colour values are stored in the table, which can be achieved with a predetermined length of the data values for the colour values.

6. Method according to Claim 5, characterized in that the method begins anew in the event of a change in the parameters on which the calculation is based.

7. Method according to Claim 5, characterized in that the calculation of further secondary colour values depends on the available computing power.