GB2287846A - Coding a video signal using constant and non-constant luminance principles - Google Patents

Abstract

In a television camera, for example, linear colour components R, G, B signals are generated by a colour correction matrix 2 from camera colour component signals r, g, b. The linear colour component signals are passed through a non-linear gamma processor 4 and a coder 6 to form conventional, non-constant-luminance luminance and colour-difference signals Y', U', V'. The colour-difference signals are low-pass filtered by a channel filter 8 for transmission. The linear colour component signals R, G, B are also passed through a coder and a gamma processor 12 to form a non-linear constant-luminance luminance signal YCL'. The non-constant-luminance and constant luminance luminance signals Y' and YCL' are then respectively low-pass and high-pass filtered by filters having complementary gain-frequency characteristics, and added by an adder 18 to form a combined luminance signal for transmission or recarding. <IMAGE>

Description

Method and Apparatus for Coding A Colour Video Signal This invention relates to a method and apparatus for coding a colour video signal, for example for transmission or for recording on a recording medium.

Two known methods for coding a video signal are nonconstant-luminance (NCL) coding and constant-luminance (CL) coding. These techniques will be briefly reviewed with reference, by way of example, to television signal transmission and decoding.

Non-Constant-Luminance Codinq In conventional NCL coding the luminance signal Y' is formed from non-linear colour component signals R',G',B' using the coding equation: Y'=ryR'+gyG'+byB' where the primes indicate non-linear gamma-correction for a reference display and the multipliers ry, gy, by are the proportions found by analysis of the RGB reference display primaries:

IX rx gx bx R #Y#=#r, g, by#.#G# Z rz gz bz B The middle line of this equation defines perceived luminance Y in colorimetric terms. Clearly, the coding equation above does not give the same luminance signal value as the arithmetically correct equation derived directly from the middle line of the matrix equation given above: Yl=(ryR+g,G+b,B)' However, use of the "incorrect" equation can give correct results in practice for the following reasons. A luminance signal Y' and two colour difference signals U' and V as follows are transmitted: Y'=ryR'+gyG'+byB' ; U'=a(B'-Y'); V'=P(R'-Y') where the colour differences U' and V' are low-pass filtered by channel filters before transmission to match the colourdifference channel bandwidth of the transmission medium. At low frequencies these signals decode as follows:

However at high frequencies, where U'=V'=O, decoding gives: R'=G'=B'=Y' This may appear to be correct but analysis shows that a variable amount of the perceived luminance at the display where the signals are received is carried in the colour difference signals. For example, for pure red: R'=R=1; G'=G=O; B'=B=O and the correct perceived luminance in the image (using the centre line of the matrix equation) is: Y=ry or Y'=ry' The transmitted signals are: Y' =ry; U'=&alpha;(-ry) ; V'=ss(1-ry) which at low frequencies are decoded as follows: R' P--=1; G'=O; B'=O This is correct, but at high frequencies where U'=V'=0, decoding gives: R'=G'=8'=Y'=ry which is smaller than the correct value ry'. Evidently some luminance information (ry'-ry) travels through the colour difference channels, and is therefore attenuated at high frequencies by the channel filtering process. The loss can be dramatic, only 11% of the luminance detail in primary blue (R=o, G=O, B=l) travels through the "luminance" channel, when using the EBU primaries and transmission equations.

One disadvantage of some luminance information travelling in the colour-difference channel is noise pollution of the perceived luminance. Ideally, noise in the colour-difference signals produces only hue (phase) and saturation (amplitude) perturbation in the display, both of which are less visible than luminance disturbances. Colour-difference or chrominance noise decoded as luminance is particularly visible at low frequencies.

This is especially evident in noisy environments such as lowbandwidth recorders (e.g. VHS etc).

The visibility of the resolution loss resulting from use of NCL coding depends on the chrominance channel bandwidths. If the chrominance channel filter amplitude response has a sharp cut-off characteristic, maintained for example up to approximately 50% of luminance bandwidth, then the resolution loss is problematic only for top-quality pictures and displays.

If, however, the chrominance bandwidth is low (less than approximately 30 of luminance) or has a slow roll-off characteristic within a typical 50% bandwidth, the resolution loss can be dramatic.

Constant-Luminance Coding CL coding is based on colorimetrically correct mathematical principles. The luminance signal is thus generated according to the centre line of the matrix equation derived from rigorous analysis of the RGB display primaries:

X r, g, bx R #Y#=#ry gy by#.#G# Z rz gz bz B Gamma-correcticn is then applied to give: Y'=(ryR+gyG+byB)' Colour-difference signals are formed, and low-pass filtered, as for NCL coding: U'=a(B'-Y'); V' =P (R' -Y') On decoding at low frequencies the following equations are used:

which give correct values for R',G' and B'. The double prime indicates reciprocal gamma-correction (undoing the camera gammacorrection). However, in CL coding the luminance information travels exclusively in the luminance signal, and therefore is not attenuated in areas of high colour saturation at high frequencies. Taking the previous example of transmission of the pure red primary for which: R'=R=1; G'=G=O; B'=B=O the perceived luminance in the scene (again using the centre line of the matrix equation) is: Y=ry The transmitted luminance and colour-difference signals are then: Y' =ry'; U'=a(-r') ; VT V'=ss(l-ry' ) At low frequencies, the decoder correctly produces:

whilst at high frequencies, where U'=V'=0, it produces; R'=G'=B'=Y'=ry' which is also correct. Thus no luminance travels through the colour difference channels and there is no noise pollution of perceived luminance by colour-difference noise as in a NCL coded system.

Luminance resolution is also unaffected by the choice of chrominance channel filtering. Therefore, narrower bandwidth chrominance channels can be used than for standard NCL coding.

Alternatively, the chrominance channel filtering can be simplified to implement slow roll-off characteristics, thereby reducing the ringing which usually results from the use of sharp-cut filters.

CL coding offers several advantages over NCL coding, but perhaps most significantly it preserves luminance resolution in areas of high saturation by completely separating luminance and chrominance information. This permits accurate transformation between systems using different display primaries. A major disadvantage is the extra complexity of the decoder, which requires non-linear circuits (gamma correctors), which add considerably to the cost of a decoder. In addition, transmitted CL signals cannot be decoded satisfactorily on a 'conventional' NCL decoder of the type currently in almost universal use.

The invention is defined in the appendant independent claims, to which reference should now be made. Preferred features of the invention are defined in dependent subclaims.

In the method according to the invention a composite, or combined luminance signal is generated which comprises a lowfrequency portion derived using NCL principles and a highfrequency portion derived using CL principles.

It is known that NCL coding can be colorimetrically accurate provided that the original scene or image produces no luminance frequencies higher than the colour-difference channel bandwidth. It is the colour-difference channel bandwidth limitations that prevent conventional coding from having the good luminance resolution attributes of CL coding. In practice, for example in television transmission or recording systems, the bandwidth available for colour-difference signals is much less than that for luminance signals. Consequently it is usual in broadcasting or recording colour video signals to pass the colour-difference signals through a low-pass channel filter prior to transmission or recording in order to match the available colour-difference bandwidth. High-frequency information in the colour-difference signal, which in a NCL signal can include a significant amount of luminance information, is thus lost.

If a composite luminance signal according to the invention is provided in which CL coding is used at high-frequencies, preferably above the colour-difference bandwidth limit, then many of the high-frequency luminance resolution advantages of CL coding may be retained.

Advantageously a signal comprising such a composite luminance signal can be decoded by a conventional NCL decoder.

No modification of the NCL decoder may be necessary.

In a preferred embodiment of the invention, the composite luminance signal is provided by generating CL and NCL signals from colour-component signals from a video image and producing high-frequency and low-frequency portions of the respective luminance signals by complementary filtering. The filtered signals may then simply be added together to form the composite luminance signal.

Complementary filtering of the CL and NCL luminance signals may be carried out by matching high- and low-pass filters. Alternatively the CL signal may be subtracted from the NCL signal, the resulting difference signal filtered by a lowpass filter, and the output from the filter added to the unfiltered CL signal. The resulting composite luminance signal in the latter case is equivalent to the sum of the filtered CL and NCL signals in the former case, but advantageously does not require matched high-pass and low-pass filters. The latter arrangement may thus be easier to implement in practice.

Although the techniques for generating the composite luminance signal described above can accurately restore much of the luminance information above the colour-difference channel bandwidth limit into the luminance channel, it may be possible to further enhance the resulting displayed image. When the techniques described above are used to code image areas of saturated colour the improvement over a conventional NCL system may be considerable, but in areas where the chrominance signal changes due to large changes in image colour, the perceived contrast may be reduced. For example for a change between saturated red and yellow, a NCL luminance signal has more contrast than a CL luminance signal because in the NCL signal the luminance of the red area is too low and that of the yellow area is approximately correct. NCL coding can thus produce an artificial contrast enhancement.

In a preferred embodiment, the invention can use this enhancement. Accordingly, the low-frequency portion of the composite luminance signal is generated as before by low-pass filtering the NCL luminance signal. However the high-frequency portion is generated by high-pass filtering the CL and the NCL luminance signals, comparing their amplitudes pixel-by-pixel and using for the composite luminance signal whichever has the greatest magnitude. In analogue circuitry for example a nonadditive mixer may be used to implement this technique.

To achieve a similar result, the NCL luminance signal may be spatially high-pass filtered and its magnitude used to control the degree of replacement of the high-frequency portion of the NCL signal by the CL signal.

The invention also finds application in trans coding signals between channels of different bandwidth. For example, if a NCL signal is initially generated and transmitted with a large colour-difference channel bandwidth and subsequently received by a conventional relay apparatus and re-transmitted through a channel of lower colour-difference bandwidth, when the colour-difference signals are low-pass filtered in the relay apparatus by a channel filter in order to match the lower bandwidth prior to retransmission, the higher-frequency portion of the initially transmitted colour-difference signals is lost.

Any luminance information in that portion of the colourdifference signals is also lost. This loss can be reduced by decoding the received signals into linear colour component signals (e.g. RGB) and recoding using the method of the invention with filtering to match the outgoing colour-difference channel bandwidth. Luminance information which would otherwise have been lost is thus transferred from the colour-difference signals to the luminance signal and retained.

This transcoding technique is particularly effective if the received or incoming signal has also been generated in accordance with the invention. Similarly, a signal generated according to the invention may be more accurately decoded into linear colour component signals than a conventional NCL signal to enable accurate transcoding into different display primaries.

In a preferred embodiment the invention provides a method and an apparatus for modifying a conventional NCL coding apparatus, such as a TV camera, a telecine or a recording apparatus in order to genert a composite luminance signal.

Accordingly, the linear colour component signals generated by the conventional apparatus can then be used to generate a CL luminance signal in addition to the NCL luminance signal normally generated, and the two luminance signals combined, for example filtered and summed or mixed, as described herein to provide a composite luminance signal.

This preferred embodiment of the invention may advantageously be implemented as a kit for fitting to an existing, conventional NCL coding apparatus. In, for example, a TV studio a mix of conventional and modified cameras may even be used as the signals coded according to the invention and according to NCL principles may both be decoded by a conventional decoder.

Advantageously the invention may be implemented only at a signal source, where linear colour component signals are generated. The cost of implementation is thus incurred, for example, at a transmitter rather than a receiver. Conventional NCL decoding may be used at the receiver.

Using the invention, extra luminance resolution can thus be made available to a viewer, even when using conventional NCL display apparatus, as a function of the chrominance bandwidth; the smaller the chrominance bandwidth and/or the more smoothly the chrominance bandwidth is limited, the greater the effect.

Experiments using a computer simulation have shown that luminance resolution may be dramatically increased in a low bandwidth system such as VHS video recording. The coding method and apparatus of the invention may thus be advantageously applied to code signals for recording on a recording medium, including for example video tape or other magnetic or optical media. Such a recording, in which a composite luminance signal processed according to the invention is recorded, can then advantageously contain more luminance information than a conventional NCL recording and can advantageously be replayed and displayed on conventional NCL video apparatus. The method and apparatus of the invention can accordingly be implemented advantageously in a telecine for preparing video recordings.

Consequently, references herein to transmission should be considered to encompass recording.

The invention may also be applied to any transmission standard in which CL and NCL coding can be used, such as coloured fax transmission, analogue or digital television, high definition television (HDTV) or medical imaging.

Although the invention has been described with a view to increasing video information transfer through existing channel bandwidths, it may conversely be used to retain image quality while reducing channel bandwidths.

Specific embodiments of the invention will now be described by way of example, with reference to the drawings, in which: Figure 1 is a block diagram of a coding circuit according to a first embodiment of the invention; Figure 2 is a block diagram of a coding circuit showing a variation of the circuit of Figure 1; Figure 3 is a schematic diagram of the magnitudes of the luminance signals generated and output by the circuits of Figures 1 and 2, plotted against frequency; Figure 4 is a schematic diagram of the magnitudes of the chrominance signals output by the circuits of Figures 1 and 2, plotted against frequency; Figure 5 is a block diagram of a circuit for combining a CL luminance signal and a NCL luminance signal according to a second embodiment of the invention; Figure 6 is a block diagram of a circuit showing a variation of the circuit of Figure 5; and Figure 7 is a block diagram of a circuit according to a third embodiment of the invention for transcoding video signals.

Figure 1 shows a block diagram of a video image coding circuit embodying the invention. Colour component signals r,g,b from, for example, a TV camera or telecine, are converted to linear colour component signals R,G,B by a camera colour correction matrix processor 2. As in conventional NCL coding the linear R,G,B signals then pass through a gamma processor 4 which subjects them to a non-linear power law function, as is well known, to form R',G',B' gamma-corrected colour component signals. These are input to a channel code matrix processor 6 and combined to form a NCL luminance signal Y' and two colourdifference signals U' and V. The colour-difference signals are then filtered by a low-pass channel filter 8 to form signals UT and VT for transmission or recording, the characteristics of the channel filter 8 being selected to match the bandwidth of the transmission or recording medium.

As well as being input to the gamma processor 4, the linear R,G,B colour-component signals are input to a CL coding processor 10 to form a CL luminance signal CL, which is then processed by a gamma processor 12 to generate a gamma-corrected CL luminance signal YCLI' To generate a composite luminance signal for transmission it is necessary to combine the CL luminance signal YCLr and the NCL luminance signal Y'. In Figure 1, the NCL signal Y' is filtered by a low-pass filter 14 and the CL signal YCL by a high-pass filter 16. The outputs from the filters are summed by an adder 18 which outputs the composite luminance signal yT The gain-frequency characteristics of the low-pass filter 14 and the high-pass filter 16 are preferably matched, or complementary, so that when their outputs are added, the variation of amplitude with frequency of the composite luminance signal is correctly maintained.

Figure 2 is a block diagram of a coding circuit identical to that of Figure 1 except for the arrangement used to combine the luminance signals YCLB and Y'. In Figure 2 the CL signal YCL' is subtracted from the NCL signal Y' by a subtracter 20, and the difference signal filtered by a low-pass filter 22. The filtered difference signal is then added to the unfiltered CL luminance signal YCL' in an adder 24, the output of which forms the composite luminance signal Yt.

The implementation in Figure 2 allows the use of one lowpass filter 22 for filtering the CL and NCL luminance signals YCL' and Y' without the requirement, as in Figure 1, for matched high- and low-pass filters 14,16. The Figure 2 arrangement may thus be easier to implement in practice.

The signal processing performed by the circuits of Figure 1 and Figure 2 is as follows. In each case the colourdifference signals U' and V' are generated as conventional NCL colour-difference signals. For low frequencies of the composite luminance signal, preferably within the bandwidth of the colour difference signals UT and VT, the luminance signal is produced as a NCL luminance signal Y'. For high frequencies the luminance signal is produced as a CL signal YCL' These signals are then combined by complementary filtering and adding. Preferably, the filtering should have the same response as the colour-difference signal channel filters so that any high-frequency luminance information which is lost from the NCL colour-difference signals is replaced in the composite luminance signal by information from the CL luminance signal.

The result of this is illustrated diagrammatically in Figures 3 and 4. These are plots of luminance against frequency and chrominance against frequency respectively. In Figure 4 the chrominance signals UT and VT are shown as filtered by the channel filter 8 for transmission. Figure 3 shows the NCL luminance signal after filtering by low-pass filter 14 or 22 in the preferred case where the gain-frequency characteristic of the filter is the same as that of the channel filter 8. Figure 3 also shows the corresponding CL luminance signal after filtering by the complementary high-pass filter 16 (Figure 1) or subtraction and filtration by the low-pass filter 22 (Figure 2).

The total luminance signal yT after the NCL and CL signals have been added together is also shown, indicating the way in which the CL signal introduces luminance information into the composite signal to replace that lost from the NCL chrominance signals on channel filtering.

The equations for the circuits of Figures 1 and 2 are as follows. At low frequency the luminance and colour-difference signals are: Y'=ry1R'+gy1G'+by1B'; U' a (B' -Y'); V' = (R' -Y') These can be decoded by a conventional NCL decoder to reproduce the original non-linear signals:

The proportionalities r,,,g,,,b,, are those of the conventional decoder, whether they are derived from the display primaries or not. At high frequencies the luminance and colour-difference signals are: Y' = (ry2R I gy2G+by2B) i UI =Oi =o These signals can be decoded, also by the conventional decoder, to produce: R'=G'=B'=Y' The proportionalities ry2/gy2,by2 are those of the luminance line from the synthesis matrix for the primaries of the notional display. Thus all the high-frequency luminance information available travels through the luminance channel and so the highfrequency resolution advantage of CL coding is retained.

Since the low-frequency coding is NCL, colour-difference channel noise will still pollute the perceived luminance (since some luminance travels via the chrominance channel, noise in chrominance may be decoded as luminance noise). In all-digital studio environments this need not be a disadvantage as noise levels are generally low. However, it must be borne in mind that in colorimetric terms, noise includes all disturbances of the signal, including quantisation and rounding effects and the effects of bit-rate reduction, and not just analogue transmission noise. Thus any bit-rate reduction of the colourdifference signals may still impart some effects into perceived low-frequency luminance.

It should be noted that the CL processing for luminance as shown in Figures 1 and 2 is optional in the sense that, if it is omitted, the apparatus functions as a conventional NCL coding apparatus although the enhanced performance of the composite luminance signal is of course lost. The extra circuitry may thus be retro-fitted to existing NCL coding apparatus.

It should also be noted that the extra gamma-processor 12 may not need to match the RGB gamma-processor 4 precisely since its output signal is only used at high frequencies. Some simplification of the circuitry may thus be possible. In a digital circuit, a single ROM may even be sufficient to implement all the CL processing except for the filtering and adding or combining.

Figure 5 shows a block diagram of a circuit for implementing an alternative method for combining CL and NCL luminance signals as generated, for example, in the circuits of Figures 1 and 2.

As shown in Figure 5, a gamma-processed NCL luminance signal Y' is input to a low-pass filter 30 and to a complementary high-pass filter 32 connected in parallel, and a gamma-processed CL luminance signal YCL/ is input to a similar high-pass filter 34. Outputs from the high-pass filters 32,34 are mixed in a non-additive mixer 36 which compares pixel-bypixel the CL and NCL signals and selects the signal of greater magnitude as its output. The outputs from the mixer 36 and the low-pass filter 30 are then summed by an adder 38 to form a composite luminance signal yT for output. As in Figures 1 and 2, the high-pass and low-pass filters should preferably be matched to each other and to the corresponding colour-difference channel filters.

Figure 6 shows a block diagram of a circuit similar to that of Figure 5 but which uses only low-pass filters. It therefore avoids the need for complementary high-pass and lowpass filters and so is likely to be a more practical implementation. Corresponding components in Figure 6, are designated by the same reference numerals as in Figure 5. In Figure 6, NCL and CL luminance signals Y' and CL' are input to low-pass filters 30 and 40. The low-pass filters advantageously have the same gain-frequency characteristics. The filtered luminance outputs are then each subtracted from the corresponding unfiltered NCL or CL luminance signals in respective subtracters 42 and 44, and the difference signals mixed in a non-additive mixer 36 as described above with reference to Figure 5. The mixer output is then added by an adder 38 to the low-pass-filtered NCL luminance signal output by the filter 30 to form the composite luminance signal yT.

As described above, in an area of an image where the chrominance signal changes due to a large change in scene colour, the perceived contrast may be disadvantageously reduced by using CL coding instead of NCL coding. The circuits shown in Figures 5 and 6 retain the enhanced contrast of the NCL signal by using the non-additive mixer to form the high-frequency portion of the composite luminance signal by selecting, pixelby-pixel, from between the NCL and the CL luminance signals the signal of greater magnitude. This process assures greater perceived contrast in the image resulting from the highfrequency portion of the composite luminance signal.

Other mixing or comparative techniques may be used to combine the high-fretuency NCL and CL luminance signals in either analogue or digital circuits.

Figure 7 shows a circuit for transcoding a set of luminance and colour-difference signals between two media, such as transmission or recording media, of different bandwidths.

This is specifically applicable where signals are to be retransmitted by a relay apparatus on a medium having a lower bandwidth for the colour-difference signals than the medium on which they were received. The transcoding circuit receives a luminance signal Y' and two colour-difference signals U' and V' which may be coded either by a method according to the present invention or by known NCL coding. The colour-difference signals are input to channel filters 50 for low-pass filtering to provide colour-difference signals UT and VT matched to the bandwidth of the output transmission medium.

The received luminance and chrominance signals Y',U',V' are also input to a decoder matrix 52 for decoding to form nonlinear colour-difference signals. These are processed by a gamma-processor 54 to provide linear colour-difference signals R,G,B which are then coded by a coder 56 to generate a CL luminance signal YCL- The luminance signal YCL is input to a gamma-processor 58 to form a non-linear CL luminance signal CL' ' which is then filtered by a low-pass filter 60. The filtered output forms a first input to an adder 62.

In parallel with the generation of this CL luminance signal, the luminance signal Y' received by the transcoding circuit is filtered by a low-pass filter 64 and input to the adder 62 as a second input. The low-pass filter 64 and the high-pass filter 60 should have complementary gain-frequency characteristics.

The adder 62 adds the low-pass filtered received luminance signal and the high-pass filtered gamma-processed CL luminance signal to form a composite luminance signal yT for retransmission.

Whether the signals received by the transcoder are NCL signals or composite luminance signals coded according to the invention, when the colour-difference signals UT and VT are filtered by the channel filters 50 some high-frequency luminance information may be lost. The circuit of Figure 5 allows this information to be retained by being transferred to the transmitted luminance signal yT When the received signals Y',U',V' are decoded to form linear colour difference signals R,G,B, as much of the luminance information in the received signals as possible, whether from the luminance or colourdifference signals, is retained. The R

Claims (23)

Claims

1. A method for coding a colour video signal, in which a luminance signal comprises a low-frequency portion derived using non-constant-luminance principles and a high-frequency portion derived using constant-luminance principles.

2. A method according to claim 1, in which the low-frequency portion and the high frequency portion of the composite luminance signal are generated by complementary filtering of a non-constant-luminance (NCL) luminance signal and a constantluminance (CL) luminance signal respectively.

3. A method according to claim 2, in which the NCL luminance signal is filtered through a low-pass filter and the CL luminance signal is filtered through a high-pass filter, the low-pass and high-pass filters having complementary gainfrequency characteristics.

4. A method according to claim 2, in which the NCL luminance signal is filtered through a low-pass filter and the CL luminance signal is inverted, filtered by the low-pass filter used for filtering the NCL luminance signal or by another filter having the same gain-frequency characteristic and added to the unfiltered CL luminance signal.

5. A method according to any preceding claim, in which a colour-difference signal is generated which is matched to a colour-difference output channel having a bandwidth limit, the low-frequency portion of the composite luminance signal principally comprises frequencies lower than the bandwidth limit of the colour-difference channel and the high-frequency portion of the composite luminance signal principally comprises frequencies higher than the bandwidth limit of the colourdifference channel.

6. A method according to any preceding claim, in which the composite luminance signal comprises the sum of the lowfrequency portion and the high-frequency portion.

7. A method according to any of claims 1 to 5, in which the high-frequency portion of the composite luminance signal is generated by comparing a high-frequency portion of a luminance signal derived using constant-luminance principles and a highfrequency portion of a luminance signal derived using nonconstant-luminance principles and selecting, pixel-by-pixel, the signal of greater magnitude.

8. A method according to claim 5, in which the high frequency portion of the composite luminance signal is generated by spatially high-pass filtering and rectifying the channel filtered colour-difference signal and replacing the highfrequency portion of the NCL luminance signal with the highfrequency portion of the CL luminance signal pixel-by-pixel depending on the magnitude of the filtered and rectified colourdifference signal.

9. A method for transcoding a first colour video signal comprising a luminance signal and two colour-difference signals matched to a first channel having a first colour-difference signal bandwidth limit, into a second colour video signal matched to a second channel having a second colour-difference signal bandwidth limit, the second bandwidth limit being lower than the first, comprising the steps of; decoding the first video signal to form linear colourcomponent signals; generating a luminance signal from the linear colourcomponent signals according to constant-luminance principles; and, generating a composite luminance signal for the second colour video signal by combining a low-frequency portion of the luminance signal of the first colour video signal with a highfrequency portion of the luminance signal generated according to constant-luminance principles.

10. Apparatus for coding a colour video signal, comprising: means for generating a first luminance signal according to non-constant-luminance principles; means for generating a second luminance signal according to constant-luminance principles; and means for combining the first and second luminance signals to generate a composite luminance signal, the composite luminance signal comprising a low-frequency portion based on the first luminance signal and a high-frequency portion based on the second luminance signal.

11. Apparatus according to claim 10, comprising filter means for performing complementary filtering of the first and second luminance signals to generate low-frequency and high-frequency signal portions respectively, the combining means comprising an adding means for adding the low-frequency and high-frequency signal portions.

12. Apparatus according to claim 10, comprising filter means for forming high-frequency portions of the first and second luminance signals, in which the combining means comprises a nonadditive mixer for generating the high-frequency portion of the composite luminance signal by comparing the high-frequency portions of the first and second luminance signals and, for each pixel, selecting the signal of greater amplitude.

13. Apparatus according to any of claims 10 to 12, comprising means for generating a colour-difference signal associated with the luminance signals and matched to a colour-difference channel having a bandwidth-limit, the boundary between the low- and high-frequency portions of the composite luminance signal being approximately equal to the frequency of the bandwidth limit.

14. Apparatus according to claim 13, comprising channel filter means for filtering the colour-difference signal prior to transmission, in which the low-frequency portion of the composite luminance signal is generated by filtering the first luminance signal through a filter having approximately the same gain-frequency characteristics as the channel filter means.

15. Apparatus for generating a colour video signal, comprising: inputs for receiving colour component signals; first power law circuitry coupled to the inputs to subject the received input colour component signals to a non-linear power law function; first coding matrix means coupled to outputs of the power law circuitry to code the signals into a first luminance signal and chrominance output signals; second coding matrix means coupled to the inputs to generate a second luminance signal; second power law circuitry coupled to the output of the second coding matrix means to subject the second luminance signal to a non-linear power law function; and means for combining a low frequency portion of the first luminance signal and a high frequency portion of the second luminance signal such as to provide a luminance output signal.

16. Apparatus according to claim 15, in which the first and second power law circuitry comprise gamma correctors.

17. Apparatus for transcoding a colour video signal, comprising: inputs for receiving an input luminance signal and input chrominance signals; decoding matrix means coupled to the inputs to generate colour component signals from the received luminance/chrominance signals; inverse power law circuitry coupled to the decoding matrix means to subject the colour component signals to a first nonlinear power law function; coding matrix means coupled to the inverse power law circuitry to generate a second luminance signal; power law circuitry coupled to the coding matrix means to subject the second luminance signal to a second non-linear power law function, the first non-linear power law function being substantially the inverse of the second non-linear power law function; and means for combining a low-frequency portion of the received luminance signal and a high-frequency portion of the second luminance signal such as to provide a luminance output signal.

18. Apparatus according to claim 17, in which the power law circuitry comprises a gamma corrector.

19. Apparatus for modifying a non-constant-luminance coding apparatus, comprising; means for deriving linear colour component signals from the non-constant-luminance coding apparatus; coding matrix means for coding the colour component signals into a luminance signal according to constant-luminance principles; power law circuitry coupled to the output of the coding matrix means to subject the luminance signal to a non-linear power law function; and means for combining a high-frequency portion of the luminance signal output by the power law circuitry with a lowfrequency portion of a luminance signal coded by the nonconstant-luminance coding apparatus to provide a luminance output signal.

20. Apparatus according to claim 19, in which the power law circuitry comprises a gamma corrector.

21. A method for coding a colour video signal substantially as described herein with reference to the drawings.

22. An apparatus for generating a luminance signal substantially as described herein with reference to the drawings.

23. A recording of a colour video signal coded by a method according to any of claims 1 to 9 or 21.