WO2006120488A1

WO2006120488A1 - Signal sensing and processing system and method

Info

Publication number: WO2006120488A1
Application number: PCT/HU2006/000041
Authority: WO
Inventors: Péter FÖLDESY; Csaba Rekeczky; Tamás ROSKA; Ákos ZARÁNDY
Original assignee: Mta Számitástechnikai És Automatizálási Kutató Intézet; Analogic Computers Számitástechnikai Kft.
Priority date: 2005-05-11
Filing date: 2006-05-10
Publication date: 2006-11-16
Also published as: HUP0500482A2; HU0500482D0

Abstract

The invention is a signal sensing and processing system, comprising a two dimensional sensor arrangement (20') consisting of sensors converting electromagnetic wave intensity into electric signals, a cell array consisting of cells (1) being in connection with the sensor arrangement (20'), and a signal converter unit (42') connecting the sensor arrangement (20') and the cell array. According to the invention, the cells (1) comprise - a digital data memory (26) suitable for loading in and/or reading out data centrally and for storing data of signal sensing and/or processing for the cell, said digital data memory (26) being capable for data sharing with a specified set of surrounding cells (1), - a digital processor (23) for digital signal processing and for controlling the signal sensing and/or signal conversion process on the basis of the contents of the digital data memory (26) , and - a communication unit (24) enabling data communication of the cell (1 ). On the other hand the invention is a signal sensing and processing method implemented by the system above.

Description

SIGNAL SENSING AND PROCESSING SYSTEM AND METHOD

TECHNICAL FIELD

The invention relates to a signal sensing and processing system and method, especially to the filed of sensing and processing moving and still images, preferably for a continuous sensing and processing of very high speed (in the case of appropriate lighting, even 10,000 or more frames per second) as well as high intensity moving images, and to decision making depending on the result of processing. Such tasks frequently arise in industrial quality control, in automotive industry and in military and security technology.

BACKGROUND ART

Image processing is traditionally carried out by a system consisting of a camera, an image digitiser and a sequential digital processing unit. Such a system is generally not compact, and it may not be used for high speed image processing or adaptive image sensing.

The acceleration of image processing can be achieved by various target- oriented processors (MATROX, Phillips-Trimedia) operating in parallel and developed for special image processing. Although this results in substantial acceleration, the low band with resulting in data transmission due to a full separation of sensing and processing units leads to a substantial speed drop, and furthermore adaptive sensing may not be implemented.

Numerous cellular systems have already been described for image processing purposes. Such systems and partial solutions applicable in such systems can be found for example in US 5,699,278, US 5,355,528, US 5,449,907, US 6,271 ,785, US 6,741 ,198, US 4,314,349, US 6,148,101 , US 6,735,482, US 6,728,862 and US 4,939,642. It is a common disadvantage of prior art systems that they do not allow local adaptivity necessary in the given case for image processing activities or a change in the resolution.

DESCRIPTION OF INVENTION

It is an object of the invention to implement a system and method for continuous sensing, processing and depending on the result of processing decision making for very high speed (1000 frames per second and more), high intensity dynamic range and high speed visual events, which are compact, have a low consumption, and are suitable for implementing local adaptivity necessary for image processing and in the given case for changing the resolution. A signal sensing and processing system, comprising a two dimensional sensor arrangement consisting of sensors converting electromagnetic wave intensity into electric signals, a cell array consisting of cells being in connection with the sensor arrangement, and a signal converter unit connecting the sensor arrangement and the cell array. According to the invention, the cells comprise - a digital data memory suitable for loading in and/or reading out data centrally and for storing data of signal sensing and/or processing for the cell, said digital data memory being capable for data transmission with a specified set of surrounding cells,

- a digital processor for digital signal processing and for controlling the signal sensing and/or signal conversion process on the basis of the contents of the digital data memory , and

- a communication unit enabling data communication of the cell. According to a second aspect, the invention is a method for signal sensing and processing, in which a two dimensional sensor arrangement consisting of sensors converting electromagnetic wave intensity into electric signals, a cell array consisting of cells being in connection with the sensor arrangement and a signal transforming unit connecting the sensor arrangement and the cell array are used. The inventive method is characterised by the steps of

- assigning to each cell an identical number of one or more sensors from the sensor arrangement,

- carrying out a signal conversion operation on the signals generated by the sensors assigned to the cells,

- wherein the adjustment of the sensitivity of the sensors and/or the signal conversion are carried out in accordance with the signal sensing and/or processing associated with the cell or with a specific set of the surrounding cells, by controlling the cell assigned sensor(s) and/or the signal converter unit. The invention is based on the following ideas: - by integrating close to each other the limited number of sensor elements of the imager unit (sensor block, sensor arrangement) and the processing unit (cell, processor), the data transmission time can be reduced dramatically;

- by influencing the behaviour (sensitivity, gain) of the imager unit, the high dynamic range becomes manageable by a processor located near the sensor;

- in cooperation with the controlling processor, the sensor is able to perform a dynamic range compression in the sensing period, wherein the high dynamic range (e.g. 16-bit) view is converted into a normal dynamic range (8-bit) image in a way that the contrast conditions are retained in the image generated; - in the course of image processing, performing the operations can be carried out mostly by accessing only of a small environment of the pixel;

- large number of processors optimised to perform high computational demanding tasks of image processing primarily based on local operations can be integrated on a single chip, and hence they can achieve an extremely high computational performance;

- image analysis and preparation of decision making can be greatly accelerated by this architecture;

- the complexity of decision making processes is high, but their calculation requirement is lower, and hence they can be transferred to a separate central unit.

Consequently, the invention is such a means, which comprises a K x L number of simple processors (hereinafter a cell), which cell has a simple digital processor, a digital data memory and a communication unit. The latter is characterised in that it is able to exchange data with other units outside the cell array and with similar cells located in its finite vicinity (in a distance of row R and column Q). It is also characteristic of the digital processor that it is able to increase the processing rate with the decreasing of processing accuracy, and that it is suitable for processing not only K x L, but also i*K x j*L images. It is also characteristic of the cell array that by means of the units located at its periphery, so-called 'address events' and statistical operations can be carried out with a high speed, for the whole array. The system also comprises an image sensor arrangement of M x N components, a signal converter unit required for the analog to digital conversion of the signals coming from the sensors and a (tuning) unit influencing the pixel-by-pixel the operation (sensitivity) of sensors. Furthermore, the system also comprises a central control unit supporting the standalone operation of the system.

Regarding the mode of operation, the system is able to continuously monitor and control in a content-depending way the sensor signals. Therefore, it can carry out the restarting of the sensors while the images are being sensed, or stopping them one by one (i.e. performing sensitivity adjustment), followed by processing the sensed images in accordance with data stored in the memory of the cell and in the memory of the cells linked directly to the said cell, and on the basis of the data sent by the decision making processor in the central unit.

BRIEF DESCRIPTION OF DRAWINGS

Hereinafter, the invention will be described by means of preferred embodiments as shown in the drawings, where Fig. 1 is a schematic view of a preferred system according to the invention,

Fig. 2 is a schematic view of the operation of a cell array in the system shown in Fig. 1 ,

Fig. 3 is a schematic view of a sensor arrangement embedded in the cell array shown in Fig. 1 , Fig. 4 shows another possible alternative of the arrangement shown in Fig.

3,

Fig. 5 is a schematic view of the cells in the system shown in Fig. 1 ,

Fig. 6 is a schematic structural view of another preferred embodiment of the system according to the invention, and Fig. 7 is a schematic structural view of the cells in the system shown in Fig.

6.

MODES FOR CARRYING OUTTHE INVENTION

The system shown in Fig. 1 comprises a cell array 2 consisting of complex cells 1 arranged in a two dimensional grid. It is characteristic of the cell structure according to the invention that disregarding the boundary cell rows and cell columns they are functionally identical and they are in close connection with a sensor arrangement which converts an electromagnetic wave intensity into electric signals. The cell array 2 preferably comprises the cells 1 arranged in columns 16.

The system according to the invention and depicted as an example also comprises a central control unit 41 being in connection with the cell array 2. The central control unit 41 comprises a decision making processor 8 as a central element, a cell array control unit 6 which works on the basis of the control provided by the decision making processor 8 and which is linked to the cell array 2 and issues operating instructions to the cells 1 , an input-output unit 3 controlled by the decision making processor 8, which supplies data to and obtains data from the cells 1 , and a central communication unit 11 which is capable to communicate with external units. The decision making processor 8 has a complex design, and in addition to making the decisions it controls the whole processing process.

The control of the cell array 2 is provided by the cell array control unit 6 via a cell control instruction bus 7. The program to be executed is entered in the cell array control unit 6 through the central communication unit 11 , via data and instruction buses 12. The data and instruction buses 12 can be linked in a way known per se to external computer systems, data storage units, and microprocessor units.

The cell array control unit 6 is controlled by the central decision making processor 8 via an instruction bus 10. The cell array control unit 6 is connected to the central communication unit 11 via a data bus 14. Furthermore, the cell array control unit 6 performs by means of row/column selecting signals 19 the selection of the cells 1 in the cell array 2.

The data required for operating the cell array 2 are supplied by the input- output unit 3 via a data bus 4 in accordance with the instructions issued by the decision making processor 8, which instructions are received by the input-output unit 3 via an instruction bus 9. Furthermore, the input-output unit 3 is connected via a data bus 13 to the central communication unit 11. Hence, the data to be loaded in the input-output unit 3 and necessary for operating the cell array 2 may come from an external source, when they are sent through the data and instruction buses 12, the central communication unit 11 and via the data bus 13, or these data may also be generated by the decision making processor 8, in which case they are supplied to the input-output unit 3 via the data and instruction bus 15, the central communication unit 11 and then the data bus 13. The results generated by the cell array 2 are supplied to an external store via the data bus 4, the input- output unit 3, the data bus 13, and the central communication unit 11 via the data and instruction buses 12, or - via the data and instruction bus 15 - they may serve as an input for the decision making processor 8.

The operating mechanism of the system shown as an example in Fig. 1 can be the following. The decision making processor 8 loads a series of instructions into the cell array control unit 6, and then from an external source it loads data into the cell array 2 via the input-output unit 3. Then, the cell array control unit 6 controls the operation of the cell array 2. On the basis of a control sequence implemented on the cell array control unit 6, the cell array 2 may process data and images in a way known per se, may sense and store images, and furthermore according to the invention on a cell level it may modify autonomously the sensitivity and gain of the sensors on the basis of the view sensed. On the basis of the program of the decision making processor 8, it may evaluate the results generated by the cell array 2, may make decisions and may communicate to the outside world these decisions and/or results.

Fig. 2 shows a schematic view of controlling the cells 1 in the cell array 2. According to the invention, preferably in a given moment of time each cell 1 receives identical instructions, i.e. the cells 1 are controlled in parallel and collectively, on the basis of a program stored by a control unit outside the arrangement. Preferably, data transmission is carried out in columns (or rows).

In a way shown in Fig. 2, the data bus 4 proceeds column by column in the cell array 2, and is connected to each cell 1 by data lines 5. The cell control instruction bus 7 also proceeds column by column in the cell array 2. For selecting a specified cell or cells 1 , row/column selector signals 19 provided by the cell array control unit 6 are used.

According to the invention, the electromagnetic wave signals - preferably of light having visible or non-visible wave length - are detected by a sensor arrangement 20, which is preferably also arranged in a two dimensional grid. For uniform detection, it is advisable if the elementary sensors 17 in the sensor arrangement 20 are arranged as uniformly as possible in the two dimensional grid. The sensors 17 can have identical structure, but according to any certain preferred pattern they may also have structures different from each other.

The cell array 2 and the sensor arrangement 20 are preferably located close to each other due to the advantages of tight communication requirements between them. According to a possible and extremely preferred embodiment, the cell array 2 and the sensor arrangement 20 are located physically in the same two dimensional grid, i.e. embedded into each other. The physical design of the arrangements in this case can be for example that the cells 1 are located in places remaining between the sensors 17. In the given case - provided that an appropriate technology is available - it may also be conceived that the sensor arrangement 20 is formed on an upper layer, and the associated cell array 2 on a layer therebelow.

Another possible way of the interrelated location of the cell array 2 and the sensor arrangement 20 is when the cell array 2 and the sensor arrangement 34 are arranged in physically separated two dimensional grids. In this case it is advantageous that the sensor surface can be covered by the sensors 17 with a high fill factor, because the places between them are not to be reserved for the cells 1. It can be disadvantageous, however, that in this case the space requirement is larger and that there are slightly longer signal path routes between the arrangements. Furthermore, in the two cases above, the design of the signal converter unit 42 digitizing the signals of the sensor arrangements 20, 34 for the cell array 2 will be different as well, as described below.

In Figs. 1 to 5 an embodiment is described, in which the cell array 2 and the sensor arrangement 20 are located in a physically embedded way, mapped into each other. Fig. 3 shows a cell/sensor arrangement used by way of example. According to the invention, preferably several sensors 17 are assigned to one cell 1. The number of sensors 17 assigned to one cell 1 is x*y in Fig. 3. It can be seen in the schematic view that within the physical design of the cell 1 indicated by a dotted line, the elementary sensors 17 are located in a way that they ensure the most uniform sampling of the sensed image.

The advantage of the arrangement in which several sensors 17 are fitted within a single cell is a scalable image resolution, which means that this unit is able to provide both low and high resolution sensing. As an example, the low resolution sensing can be done in a way that the signals of the sensors 17i,i ... 17_X|y assigned to one cell are averaged, and the signal processing tasks are carried out by the cell 1 using this averaged signal. Another possible method of low resolution processing could be that one of the assigned sensors 17 for example the sensor 17-ι,i is linked to each cell 1. A further possible case is when not one but the signals of several sensors, i.e. the signals generated by a sub-set of the assigned sensors are processed. Based on the description above, the advantage is that in the case of low resolution processing, the calculation time decreases proportionally with the reduction of number of pixels. In the case of high resolution image processing, the processing task can be carried out for each sensor 17. Due to the fact that according to the invention the cells 1 are preferably controlled in parallel and collectively, the signal processing covering each sensor 17 is preferably performed in a way that in each cell at a certain time the data from the sensor 17 position of an identical index is processed, after which a switching is carried out in the next step to the sensor 17 position of the next index in each cell. According to the invention, of course any arbitrary transitions can be implemented between the low and high resolution processing, by the programmable assignment of the sensors 17 to the cells 1.

Fig. 4 shows another cell/sensor arrangement by way of example. Of course, arrangements other than those depicted are also possible. In the given case a system may also be implemented, in which a single sensor is assigned to one cell. In this case of course the advantages offered by scaling are non available.

Fig. 5 shows a schematic view of a cell 1 , which can be applied in the case of a cell array 2 and a sensor arrangement 20 embedded into each other. A specified part 20' of the sensor arrangement 20 is assigned to the cell 1. The signals 30 of one or more sensor(s) 17 located in the part 20' are supplied to a signal conditioning, signal selecting and sampling unit 21. The control of the sensors 17 assigned to the cell 1 is preferably implemented also by the latter unit 21. A digitising unit 22 is coupled to the unit 21 , and from the said digitising unit 22 the data are supplied to a communication unit 24 via a digital signal bus 33.

The digital data memory 26 is linked to the communication unit 24 via a data and control bus 29. According to the invention, the digital data memory 26 can be accessed by the input-output unit 3, and it is suitable for storing the captured pixel values as well as the processing results calculated by the system according to the invention, and in the way described below, it can share data with a specified set of the surrounding cells 1. Furthermore, via the data bus 28, a low complexity, general purpose digital processor 23 is connected to the communication unit 24. On the basis of signal sensing and/or processing implemented in the cell 1 or in a specified set of the surrounding cells 1 or on the basis of centrally loaded data, the digital processor 23 controls the sensors 17 assigned to the cell 1 and/or the unit 21 and furthermore carries out processing the data (images) stored in the digital data memories 26. The communication unit 24 implements the data flow of the cell 1 in a way that through the data lines 5 it enables the connection to the data bus 4, and furthermore via the data lines 27 the connection of a specified set of the surrounding cells 1 to the communication units 24. Consequently, the cells 1 communicate with the other cells 1 in their specified vicinity, in a way that they have direct access to the digital data memory 26 of these cells, without slowing down the operation of the involved cells.

Preferably, the digital processor 23 comprises two major components: an arithmetic processor (ALU) and a morphologic processor. The arithmetic processor is responsible for the processing of greyscale images, while the morphological processor handles binary images.

In a preferred arrangement, the arithmetic processor can comprise a shift register, an accumulator, flags, saturation logic, as well as adding, subtracting and optionally a multiplying unit as well. By means of the arithmetic processor, the following operations can be carried out preferably: summing, subtracting, calculation of absolute value, comparison and in the given case multiply- accumulate.

The morphologic processor has been designed for performing logical operations. The characteristic of this unit is that in one step it can carry out logical operations even on several pixels. The operations that can be carried out by the unit also comprise maskable local and spatial logic operations. The application of the digital processor 23 comprising the arithmetic and morphologic options, enables the processing of both greyscale and binary images with optimal operations adjusted to the type of image.

The specified set of the surrounding cells 1 could be for example only some or all the neighbouring four to eight cells, but this scope may also be wider and cover a larger neighbourhood of the cell 1. Taking into consideration the sensed pixel values of the neighbouring cells 1 is necessary, because most of the image processing operations - in a way known per se - not only takes into consideration the given pixel, but also the values of the surrounding pixels. The operation of the cell 1 according to Fig. 5 can be carried out as an example as follows. In accordance with the instruction bus 7, the digital data memory 26 is loaded with data from the input-output unit 3 and/or from the digital data memories 26 of the surrounding cells 1. Next, the digital processor 23 starts via the instruction bus 25 the sensing of the part 20' of the sensor arrangement 20, and thereafter carries out a sampling by means of unit 21. Next, the signals of the unit 21 are digitised by the digitising unit 22 and the digitised values are stored in the digital data memory 26. The so measured and stored signals of the sensors 17 can be processed by the digital processor 23. Depending on the result, when taking the following image, the digital processor 23 is able to operate the given sensor or a specified set of the sensors in accordance with the sensed light intensity differently from the other sensors. As a result, a locally adaptive image sensing can be implemented, by which in the case of large dynamic differences in a given sensing area, the sensitivity of the sensors of bright areas is reduced. Decreasing the sensitivity can be implemented for example by controlling the sensitivity of the sensors assigned directly to the cell, or in the given case this can be achieved also by appropriately controlling the gain of the unit 21.

It can be seen that the cells 1 can be considered to be so-called complex state machines, which can execute instructions differently from each other on the basis of the instructions and data received from the central control unit and furthermore by the complex combination of own data and the data accessed in its vicinity.

Naturally, the system according to the invention, may be extended with preferred approaches already known from prior art, for example with the so-called micro-lens structure. The micro-lens structure means a grid arrangement consisting of micro-lenses placed above each sensor, and this expands the sensing area beyond the sensors 17 of limited surface by directing the light beams - originally falling between the sensors 17 - to the sensors 17. Fig. 6 shows a further preferred embodiment of the system according to the invention, in which the cell array 2 and the sensor arrangement 34 are designed in grids physically separated from each other. In this embodiment, the signal converter unit 42 transforming for the cell array 2 the signals of the sensor arrangement 34 appears as a separate unit, which signal converter unit 42 is implemented in parts 42' in the embodiment according to Figs. 1 to 5, as shown also in Fig. 5. In this embodiment, the signal converter unit 42 comprises a shutter controlling and A/D unit 35 connected to the separate sensor arrangement 34 and a digital multiplexer 36 connected to unit 35. The unit 35 is linked to the sensor arrangement 34 by a parallel analogue signal bus 37. This unit performs the control of the sensing by the sensors and of signal conversion, and carries out analogue-digital conversion of the sensed signals. Via a parallel digital signal bus 38, the unit 35 is connected to the digital multiplexer 36. Via a data bus 39, the multiplexer 36 communicates with the input-output unit 3. The multiplexer 36 carries out signal selecting tasks known per se. Consequently, the sensed signals reach the cell array 2 via the signal converter unit 42. The adaptive shutter control is performed via the same route, but in the opposite direction.

Fig. 7 shows a schematic view of the cells 1 used in the system according to Fig. 6. Since in this embodiment the sensor arrangement 34 is formed in a separate grid, the cell 1 does not comprise the unit 21 and the digitising unit 22, but the data necessary for operation are supplied to the cell 1 via the data line 5, from outside the cell.

In the embodiment according to Figs. 6 and 7, the A/D conversion used in the unit 35 is preferably implemented by a single row (or column) consisting of A/D converters, in which the number of converters is adjusted to the number of rows or columns in the sensor arrangement or in the given case to the number of rows and columns in the cell array. Consequently, the A/D conversion can be carried out preferably row by row or column by column. In the embodiment according to Figs. 1 to 5, on the contrary, a separate A/D converter is to be arranged into each cell. From the sensor arrangement 34, an image with variable resolution can be cut out during the execution of the program from arbitrary location. The cut out position might defined by the image content. The image supplied to the cell array 2 enables among others by the system according to the invention, a biologically motivated sensing method similar to the foveal vision known from human vision. Foveal vision means that on a relatively large sensed area a field of vision - the fovea - representing a partial area thereof is moved, and the fovea is directed to or swept over appropriate points in order to sense accurately the details important to us. Those parts of the sensing area which does not contain important information for us are not scanned by the fovea and hence significant signal sensing and processing capacity is reserved.

By implementing the foveal sensing according to the invention, it can be achieved that the system is made to function like sensing a high resolution and sharp image in the total sensing area, because the location of the fovea - the cell assignment to a specified range of sensors - can be modified arbitrarily according to areas significant from the aspect of acquiring information.

The digital processor 23 of the cells 1 in the system according to the invention can be implemented preferably by a bit serial architecture, which enables a relatively high density implementation and by reducing the resolution according to the description below, it also allows the flexible reduction of the calculation time. Preferably, the resolution is determined by the bit depth applied in the bit serial processor. By reducing the bit depth, the processing time is decreased linearly and in a multiplication it is reduced squarely. If the higher resolution is not necessary and for example instead of eight we calculate with a four bit depth, the processing time is reduced. As a result of the bit serial architecture, preferably the arithmetic of the digital processor 23 is of the bit serial type, the digital data memory 26 has a one bit access and data exchange with the neighbouring cells is performed bitwise.

Another preferred embodiment of the digital processor 23 of the cells 1 in the system according to the invention is if less processor but having a higher complexity - for example 8 to 16 bit - is integrated into the system. In such a case a larger number of pixels may share one processor because the processor is faster. The advantage of this solution is that it occupies a smaller silicon surface, because the control lines are to be routed to a lower number of locations, and the infrastructure which appears regardless of the complexity around each processor is to be built fewer times.

The system according to the invention can be used preferably for example to perform so-called 'address event' operations, meaning a position read-out depending on the contents. If for example we manage to shrink by an algorithmic process the image of moving objects detected on the sensing area to one or a few pixels each, then the coordinates of these points can be read by the address event process without reading the whole image, i.e. in a much shorter time. It is one of the most important advantages of the system and method according to the invention that by means of coupling the digital processors and sensors in the cells, the sensitivity of the sensors can be controlled subject to local illumination in each pixel autonomously. Hence, the signal of each sensor can be amplified and sampled regardless of each other or the sensors can be separately switched to a reset state and restarted. The controlling of the sensitivity preferably means for example regulating the integration time of the sensor signal. The integration time can be changed inversely to the illumination and it can be increased or decreased subject to the velocity of the object on the sensing area. In the case of faster moving detected objects, for example, the integration time can be reduced in order to decrease the blurring and therefore to increase the accuracy of defining the position. In the case of embedded cell and sensor arrangements, this control can be preferably implemented via the unit 21 and the instruction bus 25 leading to the sensors, and in the case of separate cell and sensor arrangements by controlling the unit 35. These units receive the command from the digital processor 23 for continuing or stopping the integration in each cell. This decision is made by the digital processor 23 on the basis of the previous image(s), the actual image, the information sensed locally or in the neighbourhood, and according to the instructions issued centrally on the basis of the data appearing in the digital data memory 26. This method called shutter control in the special field has been used in the prior art so far simultaneously and uniformly to cover a whole sensing area (for example in digital cameras). Preferably, shutter control - as known in the art - is implemented as a transistor for switching the current path between the integrating capacitance and a photo- diode generating the photo current. The input of the transistor switch is preferably represented by a 1-bit memory, the value of which is set by the digital processor 23.

As already discussed above, in the cell array 2 according to the invention, the cells 1 are controlled preferably in a parallel and collective way (single instruction - multiple data, SIMD). The initial data used for calculations and signal processing is of course locally different. The SIMD operation can be altered by the so-called maskability, which can be applied also in the system according to the invention. Masking means primarily that the system according to the invention carries out operations subject to an image content, i.e. subject to an area. A system executing the given instruction(s) on a certain area only can be suitable for example for modifying not only the integration time in a weaker illuminated area, but also for performing other processing operations in order to tune for example the contrast. Preferably, masking can be implemented by a single bit appointed in the digital data memory 26, the centrally or even locally adjusted value of which indicates whether the given instruction is to be carried out by a particular digital processor 23. Masking may be applied furthermore also for data loading for a given area.

In the system according to the invention, destructive or non-destructive sensor reading methods can be preferably implemented, and they can even be used in a programmable or selective way. Destructive read-out means that the charge is removed from the sensor appearing as a capacity at the time of readout, as a result of which read-out can only happen once after the sensing. In case of a non-destructive readout, the charge is not removed from the sensor during the read-out. It can be implemented for example on a way that the value is read out through a high input resistance sampling unit without interrupting the integration. This enables a so-called intra-frame local integration time control method, meaning that during the recording of the actual image, the calculation and the tuning of the integration time becomes possible on pixel level. The system according to the invention preferably has prior art saturation arithmetics. These arithmetics represent protection against overflowing, i.e. in a hexadecimal system, for example the result of the operation FFFE+3 will not be 0002, but FFFF. It is advisable to support the saturation arithmetics on a hardware basis by means of the application of overflow and underflow bits.

The system according to the invention can be used preferably for implementing statistical evaluations, which are typically not local operations. The statistical assessment generally represents the calculation of a single scalar result on the basis of the sensed image in a given sensing area or its processed version. Such a result can be, for example, the number of black dots appearing in the sensed image in the sensing area. The statistical analysis is preferably implemented as a linear array of cells formed as a separate row or as the bottom row of the arrangement, into which the values of the image are shifted down parallel. In each shifting step, logic or arithmetic accumulation operations can be carried out on a row-wise basis on the pixel values. When all the rows are shifted down and processed, the result can be shifted out from this statistical row, and further evaluated. The system according to the invention is preferably implemented using

CMOS technology, as a monolithic integrated circuit unit on a single semiconductor carrier (substrate) or on more than one carriers arranged side by side.

The system and method according to the invention are not limited of course to the embodiments described in details, but further modifications and variations are possible within the scope of the following claims.

Claims

1. A signal sensing and processing system, comprising a two dimensional sensor arrangement consisting of sensors converting electromagnetic wave intensity into electric signals, a cell array consisting of cells being in connection with the sensor arrangement, and a signal converter unit connecting the sensor arrangement and the cell array, c h a r a c t e r i s e d in that the cells (1) comprise

- a digital data memory (26) suitable for loading in and/or reading out data centrally and for storing data of signal sensing and/or processing for the cell, said digital data memory (26) being capable for data transmission with a specified set of surrounding cells (1),

- a digital processor (23) for digital signal processing and for controlling the signal sensing and/or signal conversion process on the basis of the contents of the digital data memory (26) , and

- a communication unit (24) enabling data communication of the cell (1).

2. The system according to claim 1 , characterised by having a central control unit (41) comprising a central decision making processor (8), a cell array control unit (6) operating on the basis of a control of the central decision making processor (8), said cell array control unit (6) being connected to the cell array (2) and supplying instructions to the cells (1), an input-output unit (3) controlled by the decision making processor (8) and supplying data to and obtaining data from the cells (1), and a central communication unit (11) suitable for communication with external units.

3. The system according to claim 2, characterised in that in the cell (1) the digital processor (23) controls the sensitivity of the sensor(s) (17) associated with the cell (1) or the gain of the signal converter unit (42), and is suitable for processing the data stored in the cell memory (26).

4. The system according to claim 3, characterised in that the cells (1) of the cell array (2) are controlled in parallel and collectively.

5. The system according to claim 4, characterised in that the signal converter unit (42) comprises a signal conditioning means, a sampling means and a digitising means, and in case the number of sensors (17) is higher than that of the cells (1), means for selecting the signal of the sensor(s) (17) to be actually connected to the cell (1).

6. The system according to claim 5, characterised in that the cell array (2) and the sensor arrangement (20, 34) are embedded into each other and the signal converter unit (42) comprises for each cell (1) a signal conditioning, signal selecting and sampling unit (21), and a digitising unit (22) connected to this unit (21).

7. The system according to claim 5, characterised in that the cell array (2) and the sensor arrangement (34) are arranged in physically separated grids, and the signal converter unit (42) comprises a shutter controlling and A/D unit (35) connected to the separate sensor arrangement (34) and a digital multiplexer (36) connected to the shutter controlling and A/D unit (35).

8. The system according to claim 5, characterised in that more than one sensor (17) is assigned to one cell (1) and the system enables a lower resolution processing by averaging the signals of the sensors (17) assigned to one cell (1) or by processing only one set of these signals, and a higher resolution processing by processing the signal of each sensor (17).

9. The system according to claim 5, characterised in that the number of sensors (17) in the sensor arrangement (20, 34) is higher than that of the cells (1), and by controlling the signal converter unit (42), each sensor (17) in a variable sub-set of the sensors (17) is connected to one cell (1) of the cell array (2).

10. The system according to any of the claims 1 to 9, characterised in that the digital processor (23) comprises an arithmetic processor for processing grey scale images and a morphological processor for processing binary images.

11. A method for signal sensing and processing, in which a two dimensional sensor arrangement consisting of sensors converting electromagnetic wave intensity into electric signals, a cell array consisting of cells being in connection with the sensor arrangement and a signal transforming unit connecting the sensor arrangement and the cell array are used, c h a r a c t e r i s e d by the steps of

- assigning to each cell (1) an identical number of one or more sensors (17) from the sensor arrangement (20, 34), - carrying out a signal conversion operation on the signals generated by the sensors (17) assigned to the cells (1),

- wherein the adjustment of the sensitivity of the sensors (17) and/or the signal conversion are carried out in accordance with the signal sensing and/or processing associated with the cell (1) or with a specific set of the surrounding cells (1), by controlling the cell assigned sensor(s) (17) and/or the signal converter unit (42).

12. The method according to claim 11 , characterised in that the cells (1) of the cell array (2) are controlled in parallel and collectively.

13. The method according to claim 12, characterised in that more than one sensor (17) is assigned to one cell (1), and a processing of a lower resolution image is carried out by averaging the signals of the sensors (17) associated with one cell (1) or by selecting the signal of only one or some sensors (17) per cell (1), and/or processing of a higher resolution image is carried out by selecting the signal of each sensor (17).

14. The method according to claim 11, characterised in that the sensitivity of sensor(s) (17) assigned to the cell (1) and/or the gain of the signal converter unit (42) are controlled on the basis of signal sensing and/or processing in association with the cell (1) or with a specified set of the surrounding cells (1).

15. The method according to claim 11 , characterised in that more than one sensor (17) is assigned to one cell (1) and by controlling the signal converter unit (42), sensors (17) in an alterable sub-set of the sensors (17) are connected to each cell (1) of the cell array (2).