GB2227899A - Colour video signal processing - Google Patents

Abstract

Colour video signals are processed or stored in chrominance component form using in addition to the U and V component signals an additional W component signal formed from G - Y. Reconversion to R, G, B is simplified as no multiplication or division operations are necessary, only simple addition of the luminance component Y. Also removal of any illegal colour values generated during processing is facilitated. <IMAGE>

Description

COLOUR VIDEO SIGNAL PROCESSING BACKGROUND OF THE INVENTION This invention relates to colour video signal processing.

In colour video systems, the video information can be represented by three colour component signals, known as the R (red), G (green) and B (blue) signals, or in a so-called encoded form. In encoded form the signals are suitable for combining into a composite signal such as a PAL or NTSC signal, such as is used for broadcast television transmission. PAL-type encoded signals comprise a Y or luminance signal and U and V signals which carry the colour information and are termed the chrominance signals. Such Y, U and V signals may constitute a studio standard for video signal processing in composite component form.

The luminance and chrominance component signals Y, U and V are derived from the R, G and B signals as follows: Y = 0.59 G + 0.30 R + 0.11 B (1) U = B - Y (2) V = R - Y (3) It will be noted that R, G, and B, and also Y, are unsigned values (i.e. do not become negative), whereas U and V are signed values. In digital signal processing it is convenient to allow, R, G, B and hence Y to range between 0 and 255.

As a general principle, signal processing can be done with the signals in R,G,B component form or in Y,U,V luminance/chrominance form. Ideally processing is done in RGB form with all three signals held to eight-bit accuracy. However, it is advantageous som-etimes to operate with the signals in Y,U,V form either because they already exist in that form or because the U and V signals can be stored at a lower resolution than the Y component, i.e. in a digital system using a lower sampling rate, without substantive impairment to the image, thus reducing the cost of the memory and processing circuitry required.

Of course, there are many instances where it is required to transform between R,G,B format and Y,U,V format. One instance where it is convenient to be able to do this is where it is desired to "recolour" a black and white still image, perhaps by hand in a graphics system. This may be referred to as "tinting" an image.

In the tinting operation, the signal is stored in digital Y,U,V form in a store known as a frame buffer, and the Y component in the frame buffer is left unchanged while the U and V components can be "painted" with new values which either totally replace or modify the old values. There will be other instances where it is desired to process only the luminance (luma) component or the chrominance (chroma) components. In general, the Y,U,V signals will typically have originally been derived from R, G, B signals and it may be desired to produce R, G, B output signals. Such an arrangement is illustrated in Figure 1 of the drawings, for the general situation where individual processing of luminance and/or chrominance is required.

The circuit of Figure 1 includes inputs lOR, lOG, lOB connected to a matrix 12 labelled matrix A which produces signals Y, U and V at outputs 14Y, 14U and 14V in accordance with equations (1), (2) and (3) above. Equation (1) may conveniently be implemented by a look-up tables (LUTs) and an adder. Processing circuits 16 and 18 operate on the luminance signal Y and on the two chrominance signals U, V to produce signals Y' and U', V' respectively. An inverse matrix 20 labelled matrix B reconverts Y', U', V' to R', G', B' which it supplies on outputs 22R, 22G and 22B.

In the output matrix 20, the mathematical operations required are as follows: R' = V' + Y' (4) G' = (Y' - 0.30R' - O.llB')/0.59 (5) B' = U' + Y' (6) The generation of R' and B' is easy but the generation of G' requires relatively complex processing on a pixel-by-pixel process, i.e. for each picture element in the image. In this processing both time and accuracy are lost. This is a particular problem because the eye is most sensitive to green, and green is the major contributor to the luminance of the signal.

SUMMARY OF THE INVENTION We have appreciated that speed and accuracy can be increased at the cost of some additional storage and processing capacity, by processing or storing not just the U and V chrominance components, but also what we term a W chrominance component derived as: W=G-Y (7) More generally, the invention is defined in the appended claims to which reference should now be made.

BRIEF DESCRIPTION OF THE DRAWINGS The invention will be described in more detail by way of example with reference to the áccompanying drawings, in which: Figure 1 (referred to above) is a block diagram of a known video signal processing system; Figure 2 is a block diagram of a video signal processing system embodying the invention; Figure 3 is a block diagram of the system arranged for a tinting operation; Figure 4 is a block diagram of the system arranged for scaling saturation; Figure 5 is a block diagram of a preferred form of the system using look-up table devices; Figure 6 illustrates the circuitry associated with one of the look-up table devices of Figure 5; Figure 7 illustrates a possible implementation of the matrix C of Figure 3; Figure 8 illustrates a version of the matrix providing increased accuracy;; Figure 9 is a block diagram of one form for the matrix D in Figure 3; Figure 10 illustrates one of the 16-bit adders of Figure 9; Figure 11 illustrates circuitry for stopping illegal values of R, G and B being produced, by clipping: and Figure 12 is a flow chart illustrating a software method of stopping illegal colours being produced using a scaling function.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Figure 2 shows a system based on the video processing system of Figure 1 but adapted in accordance with this invention. To this end, the matrix 12 includes an additional output 14W. The signals generated at the outputs 14Y, 14U, 14V and 14W are thus related to the input signals R, G and B as follows: Y = 0.59 G + 0.30 R + 0.11 B ..... (1) U = B - Y (2) V=R-Y . . . . (3) W=G-Y (7) The processing circuit 18 operates with the three chrominance signals U, V and W to produce three processed signals U', V' and W'.

Of course, there is some redundancy in the information contained in these signals. The operation of the output matrix 20 Is however much simplified as follows: R' = V' + Y' (4) G' = W' + Y' (8) B' = U' + Y' (6) The relatively complex processing of equation (5) has been replaced by the much simpler addition of equation (8).

It will be understood by those skilled in the art that in both the known system of Figure 1 and the system of Figure 2 the chrominance signals applied to and produced by the chrominance processing circuit 18 will need to be held to one bit greater accuracy than the R, G and B signals if rounding errors are to be minimised. For example, the chrominance signals may be held as 9 bit signals.

Figure 3 shows more precisely the configuration of the system in a tinting operation where an image is being totally recoloured (or a black-and-white image is being coloured for the first time).

In this instance only the luminance output of the input matrix 12 is used, and much of the matrix, labelled matrix C, can therefore be discarded. The luminance processor 16 is replaced by a direct connection and the chrominance processor takes the form of a set of three value generators 28U, 28V, 28W. The values U', V' and W' are substituted for the values of U, V and W. The signals Y',U',V' and W' are transformed into R',G' and B' by matrix 20, here labelled matrix D, in accordance with equations (4), (8) and (6). The matrix thus consists in principle simply of three adders (see Figure 9 described below).

Figure 4 illustrates the configuration of the system if the saturation of the colours in the image is to be changed. In this case the luminance processor 16 is again omitted. The chrominance processor 18 now comprises three scaling circuits 30U, 30V and 30W each of which receives the output of a scale factor generator 32.

The scaling circuits 30 may each comprise a multiplier capable of multiplying a signed nine-bit quantity by a scale factor which could also be known to 9-bit accuracy. A scale factor of 256 can conveniently be taken to provide an overall unity multiplication, and the least significant 8 bits of the resultant 18-bit product discarded to perform an effective division by 256. This would enable scale factor multiplication of between 0 and 2 to be used to an accuracy of one part in 256. The remaining ten bits of data are passed on for further processing and eventually for conversion to RGB form in matrix D. If the signal channels are each 9 bits wide, the top bit has to be discarded. This top bit can not contain data if the saturation is being decreased, as the multiplication factor from generator 32 is then less than 256.If the multiplication factor is greater than 256, however, care must be taken to ensure that the incoming saturation is sufficiently low such that overflow can not occur in the multiplier. This problem can be entirely avoided by using 10-bit channels.

A multiplier 30Y (not shown) could be included in the luminance channel in addition to or instead of the multipliers 30U, 30V, 30W for the chrominance components, to achieve peak level adjustment of the signals.

In Figure 5 is shown an alternative and preferred implementation for saturation and/or hue changes, using 9-bit addressed look up tables (LUTs) 34U, 34V, 34W. Each look up table circuit 34 is programmed to produce the desired multiplication factor of the 9-bit signed input, and to provide at least 9 bits of output. The LUT outputs may be of greater accuracy than the inputs, e.g. to 12 bits, with the subsequent processing also being done to 12-bit accuracy.

The implementation of Figure 5 is preferred because all functions of the input data may be generated, not just uniform hue and saturation changes. Thus additions may be programmed to produce overall colour casts or to correct overall colour casts.

Also the multiplication operations described can be implemented with automatic clipping to 9-bit values, to avoid the overflow problem noted above.

The look up tables are programmed by a microprocessor 36 (Figure 6) which assembles the 512 values required in each case.

They may be written into the tables at any convenient time when the circuity illustrated is not in use, such as during frame blanking in a real time system. Appropriate buffer circuits (38) are provided enabled by a frame blanking signal. All this is illustrated in Figure 6, which shows somewhat diagrammatically the circuitry associated with one LUT 34. The frame blanking signal disables the address input to the LUT by means of inverter 39 and enables the processor address bus onto the addressing of the LUTs and the processor data bus onto the data lines of the LUTs (input and output functions being shown separately for convenience). An area of the microprocessor memory map is decoded to produce the write signal to the LUT and the processor may then write the data.The processor 36 must confine its accesses to the frame blanking period which can be done by feeding the frame blanking signal to an interrupt line on the microprocessor The apparatus of Figure 5 can be extended to include a further LUT 34Y in the luminance channel, shown in broken lines on the figure. With this circuit it is possible to change the luminance signal as desired, for example, enabling black level adjustment as well as peak level adjustment as is obtainable with a multiplierbased system of the type shown in Figure 4.

Figure 7 illustrates a possible form for the matrix C (12 on Figure 3) which produces a Y signal of 8 bits and U, V and W signals of 9 bits. The R, G, and B signals are applied to individual LUTs 40 which act as multipliers and the outputs of which are applied to adders 42,44 to provide a Y signal in accordance with equation (1).

Subtractors 46 then implement equations (2), (3) and (7) respectively. In practice the range of U, V and W is limited to less than the full range of 9 bits, and specifically W is limited to 8 bits. Preserving the full 9 bits through the U, V and W modification processing is however desirable so the illegal-colour handling circuits, described below, can be implemented. Each look up table (LUT) is preprogrammed to provide the correct ratios for assembling Y, i.e. 0.59 for green, 0.30 for red and 0.11 for blue.

Thus 8-bit in to 8-bit out look up table proms (programmable readonly memories) can be used. The ith entry of each LUT is as follows: green(i) = 0.59 * i; red(i) = 0.30 * i; blue(i) = 0.11 * where * indicates multiplication. In practice this matrix is implemented with the required latches to control the data flow through the matrix. This enables a real time system to be constructed.

Figure 8 shows an alternative implementation of matrix C for higher accuracy. 15-bit LUTs 40 are used and the values are passed through at 128 times the value used in the 8-bit case.

Thus the equations for the LUTs become: green(i) = 128 * 0.59 * red(i) = 128 * 0.30 * blue(i) = 128 * 0.11 * i.

The outputs produced are 15-bit for Y and 16-bit signed values for each of U, V and W.

To adjust for the greater number of bits used in the LUTs 40 and subtractors 46 appropriate "shift left 7" circuits 48R, 48G and 48B are included to shift the R, G and B signals to the "left" by 7 bits to make 15 bit signals.

Figures 9 and 10 illustrate a possible implementation for matrix D (20 in Figure 3) using 16-bit adders 60. In practice, for an 8 or 9 bit system only 10 bits of the output are required, which could be achieved with three four-bit adders. For an 8/9 bit system the top valid bit of the U, V or W signal is shown in Figure 10 as being sign extended to 16 bits. Sign extension is required regardless of the accuracy of the system if an, adder of more bits than the UVW accuracy is required. The top bits of the Y signal are padded with zeroes.

The system as described is capable of producing "illegal" R,G and B values. These are values where these components after matrix D exceed the maximum values assigned to them, or are less than zero.

Thus of the 10 bits out of Figure 10 only 8 bits (bits O to 7) are valid R, G and B data in the range 0-255. Bit 9 indicates that the value is negative; if it is not, bit 8 indicates the value exceeds 255.

Two methods may be applied to handling illegal values. It may be acceptable to clip the values to the range O - 255, in which case Figure 11 shows a suitable implementation. This produces the value 0 if a negative input is encountered and the value 255 if a number greater than this is encountered.

The red clipper circuit 64R is shown in Figure 11 and comprises an input 66 for receiving a 10-bit signal of which the eight least significant bits are applied to a buffer circuit 68. A buffer 70 holds the value 0 and a buffer 72 holds the value 255. One only of the buffers is enabled at any moment to provide a signal to the output in dependence upon the value of the top two bits of the input signal. To this end a NOR gate 74 receives both the top bits 8 and 9 and provides an output to enable buffer 68 only when they are both zero. A NOR gate 76 receives bit 8 after inversion in an inverter 78, and bit 9, and enables buffer 72 when bit 8 is one and bit 9 is zero. The enable input of buffer 70 is activated whenever bit 9 is one.

In Figure 11 the logic has the effect of the following truth table: Bit 8 Bit 9 Meaning 0 0 8-bits valid 0 1 clip to zero 1 0 clip to 255 1 1 clip to zero.

Clipping the signal however affects the luminance value as well as the overall colour balance. If it is desired to retain the correct luminance value regardless of the processing, as would be required in a "tinting" operation where a black and white image is being recoloured, it is necessary to scale down the U,V and W values in equal proportions before converting to R,G,B. Maintaining equal proportions means that the hue is preserved; the effect is merely to reduce saturation. The minimum reduction to prevent illegal colours is desired. Two methods are described below.

The flow chart of Figure 12 illustrates an iterative method of reducing U,V and W until a valid (or "legal") combination is achieved. This is achieved by a binary search and can be implemented in software.

In an initialisation step 100 the following values are set: (a) illegal~U, illegal~V and illegal~W are set to the input values U, V and W respectively (b) legal~U, legal~V and legal~W as well as iterations are all set to zero.

The process passes through step 102 which takes the average of the illegal and legal values of U, V and W respectively. In its first loop, because the legal values have been set at zero, this will mean the input values are halved. In a conversion step 104, the YUVW values are converted to RGB and a test 106 made to see if the RGB values are valid. If they are, which is likely on the first loop, then the legal values are reset in step 108 to the values at the output of step 102. However, if the RGB values are not valid, then the illegal values are reset in step 110 to the values at the output of step 102. Either way the desired valid values are bracketed by the illegal and legal values. The count of iterations is incremented in step 112 and if the maximum count is not reached (step 114) the process returns to step 102.On each iteration it will be seen that the size of the bracket between the illegal and legal values is halved.

To achieve maximum accuracy in an 8/9 bit system, 8 iterations are sufficient, therefore a value of 8 is suitable for MAX in this loop. When MAX is reached, the loop is exited and the current legal values used as the desired output (step 116). In general, in a n bit system, n iterations are sufficient. This method is ideal for a microprocessor based solution where real time is not important. Alternatively it can be implemented as part of a real time pipeline with precisely n steps.

In a second alternative method, the result of a conversion to RGB, if invalid, can be used to generate a scaling factor which can be applied to each of the U,V and W signals to result in valid RGB values on a subsequent conversion. The scaling factor is calculated as follows: if red > 255 then red~factor = (255-Y)/V, else if red < 0 then red~factor = if green > 255 then green factor = (255-Y)/W, else if green < O then green~factor = if blue > 255 then blue~factor = (255-Y)/U, else if blue < O then blue~factor = -Y/U.

The smallest of these factors is selected and used to scale U,V and W: U' = factor * U; V' = factor * V; W' = factor * W.

Then a final conversion of Y,U',V',W' can be performed to valid R,G and B.

In order to achieve fast processing the division can be performed by passing each of the U, V and W values into a look-up table containing 256 times the reciprocal of the input number, and feeding the output of the LUT to a multiplier where it is multiplied by either Y or 255-Y according to the sign of the illegal colour.

In a system with 9 bits for U, V and W, this LUT contains only 512 entries, and it is sufficient to provide 8 bits of output as the sign bit can be ignored. This is because whenever the result is negative, the condition with the -Y must be calculated to give a positive answer. The 255-Y is calculated by a simple bit inversion.

Whichever of the above two methods is used, the operation is facilitated by the availability of the W signal in addition to the conventional U and V signals.

The methods described above are suitable for use in various applications, including a computer graphics system or in a real-time video signal processor. The use of the W signal is advantageous for signal storage and the term "processing" is to be considered as covering storage in such circumstances.

Claims (17)

1. A colour video component signal W formed by subtracting the luminance component signal Y from the green signal of the R, G, B component signals substantially in accordance with the equation: W = C - Y.

2. A method of processing a colour video signal comprising the steps of receiving colour signals in R, G, B form, converting signals into luminance and chrominance form, processing the luminance and chrominance signals, and reconverting the signals into R, G, B form, characterised in that three chrominance signals are generated, and in that the reconversion into R, G, B form comprises a simple additive function.

3. Apparatus for processing a colour video signal, comprising inputs for receiving colour component signals in R, G, B form, means for converting the input signals into luminance and chrominance form, processing means for processing the luminance and chrominance signals, and means for reconverting the processed signals into R, G, B form, characterised in that the means (12) for converting the input signals into luminance and chrominance form generates three chrominance signals (U, V, W), and the means (20) for reconverting the signals into R, G, B form comprises combining means (60) for executing a simple additive function.

4. Apparatus according to claim 3, in which the processing means comprises means for altering the tint or hue of the video signal.

5. Apparatus according to claim 3, in which the processing means comprises means for scaling the chrominance signals to alter the saturation of the video signal.

6. Apparatus according to claim 3, 4 or 5, in which the processing means comprises look-up tables.

7. Apparatus according to claim 6, including means for loading the look-up tables during video frame blanking periods.

8. Apparatus according to any of claims 3 to 7, including illegal colour removal means coupled between the processing means and the means for reconverting the processed signals into R, G, B form, and operative to modify any U, V, W signals which correrspond to invalid R, G, B values.

9. Apparatus according to claim 8, in which the illegal colour removal means operates to remove the illegal colour by an iterative process including the steps of: (i) combining the illegal U, V, W colour components with known legal colour components to generate new colour components lying therebetween; (ii) converting the new colour components to R, G, B form and (a) if the resultants are valid R, G, B values, then using the new colour components as tghe next legal colour components, or (b) if the resultantgs are invalid R, G, B values, then using the new colour components as the next illegal colour components. and repeating steps (i) and (ii) until the illegal and legal colour components are the same to a required degree of accuracy.

10. Apparatus according to claim 9, in which the steps (i) and (ii) are repreated a predetermined number of times.

11. Apparatus according to claim 8, in which the illegal colour removal means includes means for converting the colour components UY, V, W to R, G, B form and for each of the R, G, B components if the component is greater than its valid maximum, generating a factor as (1-Y)/X or if the component is negative generating a factor (-Y/X) where X is V, W and U respectively, selecting the smallest of the factors thus obtained, and multiplying (attenuating) all the components U, V and W by the selected factor.

12. A colour video signal transcoder having four inputs for receiving a luminance signal and three chrominance signals, three outputs for providing R, G and B signals, and combining means coupled between the inputs and the outputs to provide the three output signals by additive combination from the four input signals.

13. A trans coder according to claim 12, in which the combining means comprises three combining circuits each having one input connected to the luminance input and each having another input connected to a respective one of the chrominance inputs and an output connected to a respective one of the outputs.

14. A colour video signal transcoder having three inputs for receiving R, G and B signals, four outputs for providing a ] uminance signal and three different chrominance signals, and combining means coupled between the three inputs and the four outputs to provide the four output signals from the input signals.

15. A trans coder according to claim 14, in which the combining means comprises a luminance combining circuit (40, 42, 44) for forming a luminance signal Y from the R, G and B signals.

16. A trans coder according to claim 15, in which the combining means comprises three subtractors for subtracting the luminance signal Y from each of the colour components signals R, G and B.

17. A colour video processing system substantially as herein described with reference to any of Figures 2 to 12 of the drawings.