GB2113037A - Method of and apparatus for coding and decoding PAL colour television signals - Google Patents

Abstract

A PAL coder comprises inputs for Y, U and V component signals, V-axis switching circuitry for switching the V signal, and multipliers for modulating the U and V signals onto subcarrier at orthogonal phase positions. The U and V signals are added and subtracted from the Y signal and the resultant applied to a Luminance stop comb filter @. The output of the luminance stop filter is subtracted from the luminance signal to provide the encoded PAL signal. The luminance stop filter is shown in Fig. 10. A decoder using the same filter is shown in Fig. 9. The decoder/coder combination is 'transparent'; once the signal has been encoded subsequent decoding/recoding operations theoretically introduce no further impairment into the signal. <IMAGE>

Description

SPECIFICATION Method of and apparatus for coding and decoding pal colour television signals This invention relates to the coding and decoding of PAL colour television signals.

Our British Patents 1,534,268 to 1,534,270, (Inventor: Martin Weston) of which No 1,534,269 may be taken as typical, describe a digital television transmission or processing system in which there are two transmitted signals, namely a luminance signal Y2 sampled at 2fisc, and a chrominance signal C sampled at fscl where fsc is the colour subcarrierfrequency. Before digitisation the PAL chrominance signals U and V are combined into U + V on one television line and U - V on the next. The system can accept signals in separated chrominance and luminance form, i.e. YUV form, or in encoded PAL form, and deliver output signals in YUV or encoded PAL' form.Any of these signals may be in analogue form, or in digital form samples at 4fsc. Each of the conversion stages of the system incorporates an appropriate comb filter which is based on the use of a line delay device.

The transmitted signals Y2 and Cg form a package which is hereinafter referred to as W. By using the system, a country generating PAL signals can exchange programmes with a country generating YUV signals using the common signal package W. This composite signal W is such that, if it was derived from PAL, it can be transformed back into a PAL signal with, in principle, no loss. This property of no (or substantially no) loss can be referred to as "transparency". A system is transparent if the system output is essentially indistinguishable in practical terms from the system input (any overall delays being discounted).

The YUVoutput signals obtained from W, on the other hand, are impaired versions of the YUV input signals (assuming these are vertically unfiltered).

Our abovementioned patents mention that a PAL decoder can be constituted by using the PAL to W decoder connected to the circuits which convert the W components signals to Y and U, V respectively.

We have also appreciated that the route YUV-W-PAL' constitutes a PAL coder. Furthermore, as the route PAL-W-PAL' is transparent, it follows that in cascaded systems of the same type the route W-PAL'-PAL-W is also transparent. It can further be shown that in cascaded systems of the same type the route W-Y'U'V'-YUV-W is transparent. Thus it is possible, in principle, to cascade any number of systems of the same type, joined either by PAL or YUV ports without impairing the W signal package.

In particular, the route PAL-W-Y'U'V'-YUV-W-PAL is equivalent to the route PAL-W-PAL' which is, in principle transparent. Thus, decoding using the system followed by recoding also using the system is transparent. Further the route YUV-W-PAL'-PAL-W-Y'U'V' is equivalent to the route YUV-W-Y'U'V',so that coding by the system followed by decoding by the system is equivalent to a YUV transmission path in which luminance is never mixed with chrominance.

Thus it is possible to PAL code and decode using the system and eliminate luminancechrominance cross-effects completely in exchange for the impairments introduced into the YUV signals.

These impairments are a reduction in vertical resolution of chrominance and vertical resolution of highfrequency luminance, and the introduction of luminance aliasing and U-V crosstalk. Such coding and decoding is hereinafter referred to as clean coding and decoding.

In accordance with another feature of this invention, this property of eliminating cross-effects depends only on the transparency of the W-PAL'-PAL-W path and thus is independent of the nature of the filters in the YUV-W and the W-Y'U'V' paths. The filters in these paths can thus be chosen with other factors in mind.

In accordance with a further feature of the invention, the transparency of the W-PAL'-PAL-W path is maintained if the delay element in the comb filters involved in this path is any odd number of lines, e.g. 313 or 625 lines.

In accordance with yet another feature of this invention, the sampling in a PAL coder or decoder can be achieved by analogue multiplication followed by simple addition. Over the video band this provides spectrum folding in a manner identical to sub-Nyquist sampling.

Preferably in constructing a PAL coder or decoder embodying the invention, the comb filters in the path between YUV and PAL are collapsed so as to require a minimum of delay elements.

The present invention is defined in the appended claims, to which reference should now be made.

The invention will be described by way of example with reference to the drawings, in which: FIGURE 1 is a block diagram of a colour television transmission system based on our British Patent No 1,534,269; FIGURE 2 illustrates the form of the filters in the system of Figure 1; FIGURE 3 is a frequency diagram illustrating how analogue multiplication and addition are equivalent to sampling; FIGURE 4 is a block circuit diagram of a PAL coder embodying the invention; FIGURE 5 is a block circuit diagram of a PAL decoder embodying the invention; FIGURE 6 is a block circuit diagram of a rearrangement of the circuit of Figure 4; FIGURE 7 is a block circuit diagram of a rearrangement of the circuit of Figure 5; FIGURE 8 is a block circuit diagram of a further simplified rearrangement of the circuit of Figure 4;; FIGURE 9 is a block circuit diagram of a further simplified rearrangement of the circuit of Figure 5; and FIGURE 10 is a block circuit diagram showing the form of the luminance stop filter used in Figures 8 and 9.

In the system shown in Figure 1, signals can be received either in Y,U,V, form or in encoded PAL form, and can be outputted either in Y'U'V' form or encoded PAL' form. Various comb filters Fro1, F,,, F4,, F" and F2, are included at the input and comb filters Fro2, F32, F42, F,2 and F22 are correspondingly included at the output as shown. Filters F", F2, and F,2, F22 are fundamental to the system. Filters F" and R,2 provide a comb filter with a modulus sine response peaking at integral multiples of the television line frequency, while filters F2, and F22 provide a comb filter with a similar response but peaking at odd integral multiples of half the line frequency.The other filters, being those in the YUV-W and W-YUV paths and which we have now appreciated are not fundamental to the system, have responses which peak at multiples of the line frequency. It will be seen that over the regions where comb filtering takes place, the transmission/frequency characteristic of all these filters is of a modulus sine form. In the central block marked W the signals comprise a luminance signal Y2 sampled at twice the colour subcarrier frequency (2fisc) and a chrominance signal C, sampled at the subcarrier frequency (fsc) and consisting of U + V on one television line and U - V on the next. Switching circuits are accommodated in the V and V' signal paths to take account of the PAL switch.It will be seen that the input and output signals are in analogue form and hence analogue-to-digital converters (ADC's) and digital-to-analogue converters (DAC's) are included.

As mentioned above, the elimination of cross-effects in the system depends only on the transparency of the W-PAL'-PAL-W path, and this is independent of the nature of the filters Fro1, F31,a and F4, in the YUV-W path and filters Fro21 F32 and F42 in the W-Y'U'V' path. The nature of these filters does, however, govern the form of the impairments introduced into the YUV signals. Figure 2 shows the form of all the filters except F01 and F02 as described in our Patent 1,534,269, filters F01 and F2 being identical to filters F" and F,2 respectively. For further details of these filters reference should be made to our British Patent 1,534,269.In Figure 2 of this application the same reference numerals are used as in our earlier patent.

Now, the 2fsc sampling action in the luminance path from Y to W can- be accomplished, over the video band, by analogue multiplication and addition. Referring to Figure 3, the uppermost waveform shows the effect of sampling the video baseband, itself shown in the second line, at a frequency such as 2fsc which is below the Nyquist limit, i.e. is less than twice the maximum video frequency. As is well known, the action of sampling causes the spectrum to be repeated infinitely at spacings equal to the sampling frequency. Within the video band therefore, the sampled signal consists of the baseband .signal together with a signal formed by spectrum folding at the upper end of the video band. The relation of these alias components to the television line frequency is well discussed in our earlier patents.

The lowest line in Figure 3 shows what happens when the baseband signal is modulated onto a frequency of 2fisc or in other words is multipiied by it. Here the effect is not to cause an infinite number of repetitions of the video band, but rather to cause just two, at 2fisc spacing from the baseband signal.

We have appreciated that if this modulated signal is now added to the baseband signal, then over the video band the result is the same as is achieved by sampling, with spectrum folding occurring at the edges of the video band.

The sampling action at fscl low pass filtering and re-modulation in the chrominance path ofthe coder can be accomplished by straightforward modulation. In the decoder, sampling at fas of the chrominance bandpass signal is equivalent to demodulation at the appropriate phase. Thus it is possible to construct an analogue coder and decoder which produce exactly the same signals as the digital system of our earlier Patents.

Figures 4 and 5 show respectively an analogue PAL coder and decoder embodying the invention.

In Figure 4, the input luminance, filtered by filter Fro1, is mixed in an adder 42 with a version of itself which is modulated in a modulator 44 on a carrier of frequency 2fsc. Such a modulator, as described above with reference to Figure 3, serves to introduce the luminance alias component, and may be termed a modifier. The mixture of wanted and alias luminance is then filtered by filter F,2, which removes half-line frequency offset alias components, those at (n + 2)fnl generated from line-offset input components, while allowing the components which are at multiples of the line frequency to pass.

Finally, the signal is filtered by a low-pass filter LP2, which is the implicit luminance DAC filter in Figure 1, to remove components, particularly alias components, beyond the nominal video bandwidth.

The U and V chrominance signals are, as in Figure 1, assumed to be limited in bandwidth to and are prefiltered by their respective filters F3, and F41, the V signal having been PAL switched in a circuit 46. They are then combined in an adder 48 into a single chrominance signal in the form of U + V and U - V on alternate lines,and modulated in a modulator 50 onto a PAL subcarrier (fsc) whose phase is appropriately related to the modifier phase. This relationship is such that peaks of the subcarrier coincide with peaks of the twice-subcarrier (luminance) modifier frequency.The modulated chrominance is then passed through comb filter F22 which allows components which are at odd integral multiples of the line frequency to pass and restores the quadrature phasing of U and V signals in uniform-coloured areas. The modulated chrominance is then combined in an adder 52 with the luminance to give a PAL signal. In practice the video band filter LP2 can be placed after the adder 52.

In the decoder of Figure 5, the PAL signal is split into two paths by filter F" and F2,. The effect of the filter F", which averages across a one-line delay in the chrominance band and thus has peak response for line-offset components, is to co-phase the U and V modulated chrominance components in uniform-coloured areas. Thus by adding in an adder 62 a PAL-modified version from multiplier 63 of the output of F" to itself, the chrominance may be cancelled out, provided the modifier phase is appropriately chosen. The signal so processed is then filtered by F02 to reduce or eliminate, depending on Fro1, the alias spectrum generated by the modifier. Finally, the filter LP2 ensures that the Y signal is confined to the nominal video bandwidth.The effect of filter F21, which takes half the difference across a one-line delay and thus has peak response for half line frequency offset components, is also to co-phase the U and V modulated chrominance and null the line-offset luminance. Thus the output of filter F2, requires only a single demodulation in demodulator 66 after which it is low-pass filtered by filter Lip,, which is the implicit chrominance DAC filter in Figure 1. Finally the combined chrominance signal is split into U and switched V components by filters F32 and F42 and the V switch is removed by a circuit 68.

We have now found that the delay elements in the filters in the W-PAL'-PAL-W path (the PAL filters) need not be one-line delays but can be of any odd number of lines. These filters are filters F", F,2, F21, and22, and in particular they can be 313 or 625 lines long, i.e. essentially one field or one picture. Where the delay is N lines long, then the comb filter will have a modulus sine response with peaks at frequencies equal to fJN; fL being the line frequency. When longer delays are used, then the YUV output signals obtained by conventional decoding of the clean coder output may exhibit movement smearing.

However, we have also appreciated that the other comb filters, namely the filters Fro1, Fro21 F3g, F32, F41 and F42, are not as such fundamental to the coder/decoder operation although they can be chosen with a view to optimising system performance.

Filters of interest in the F01 and F02 positions are (a) where they take the form of F" and F,2 in Figure 2 with a 1 or 313 or 625 line delay element, or (b) where they are preferably equal and of any one or two dimensional low pass form, cutting near the subcarrier frequency.

Filters of special interest in the F31, F32, F41 and F42 positions are (a) where they take the form of Figure 2 with a 1 or 313 or 625 line delay element, or (b) where they take weighted contributions from any number of consecutive lines in one field to form a vertical low pass filter cutting near 78 c/ph (cycles per picture height), such that F31 and F32 are equal, apart from a scale factor, and the weighting coefficients of F41 are those of F31 except that alternate values are reversed in sign. Thus, for example, it is possible to have filters based on one-line delays in the luminance path and on 313-line delays in the chrominance path, taking advantage of the lower bandwith needed for chrominance in moving areas.If the group delays of the luminance and chrominance filters differ, then appropriate compensating delays must be inserted before or after the filters.

The Table appended to this description lists combinations of luminance and chrominance filters of interest showing the delay element of the filters and the amount of compensating delay. The latter is distributed between the coder and decoder in such a way as to minimise the vertical-temporal misregistration at the output of the coder. In the Table, fifteen cases are given and for each the first line gives the magnitude of the main delay, and the second line (or second and third lines) gives the magnitude of the compensating delays. The letter n (possibly with subscripts) is used to mean an integral number of lines, while 1 D and 2D mean respectively a filter of any one or two dimensional low pass form. A one dimensional filter is simply a low pass filter to give horizontal filtering, while a two dimensional filter includes line delays to give vertical filtering within a field.

All the combinations of the Table could, in theory, be used with 313for 625-line based filters at F11, F12, F2, and F22 (the PAL filters). However, combinations of practical interest are: (a) all the combinations of Table and line based PAL filters.

(b) 313-line based combinations of the Table and 313-line based PAL filters.

(c) 625-line based combinations of the Table and 625-line based PAL filters.

It is possible that the combinations which include 625-line based filters would lead to unacceptable movement smearing, but the one (case 1 5) with 625-line filters in all positions corresponds to perfect transmission of stationary pictures, i.e. no cross-effects and no loss of resolution in the Y, U and V signals. For this case the clean coding produces an identical signal to conventional coding (discounting any delay), apart from reducing noise, and it is the clean decoder which is of primary interest.

For moving pictures, an adaptive system could be devised which switched between 625-line and 313-line or 1-line based filters according to the decision of a motion sensor. With such an arrangement it would be necessary to ensure no difference of overall group delay between the different states.

The luminance path in the coder of Figure 4 includes two comb filters in series, F01 and Fit2. These, and the other series connected comb filters, can be rearranged so that the circuits of Figures 4 and 5 become the circuits of Figures 6 and 7. In the circuit of Figure 6, the filters F, and F2 control the wanted and alias luminance spectra respectively. Here again the technique of multiplication and addition is used to generate the required spectrum folding. In the chrominance path, it is possible to modulate the U and V filters onto subcarrier as shown and then to filter using filters F3 and F4 together with a multiplier 70 and adder 72 operating on the composite U, V signal. The filters F3 and F4 control the asymmetry of the quadrature-modulated chrominance spectra about their carriers.An equivalence exists between filters F1 and F2 and filters F01 and F12 and between Filters F3 and F4 and filters F31, F41 and F22, as discussed below.

In the decoder of Figure 7 filters F5 and F6 control the wanted and alias luminance spectra respectively, whilst discriminating, with the aid of the modifier, against the subcarrier. Filters F, and F8 control the wanted and cross-talk chrominance spectra, while discriminating against line-locked luminance. In the same way as in the coder, an equivalence exists between filters F5 and F6 and filters F11 and F02, and between filters F7 and F8 and filters F22, F32 and F42.

A problem with the arrangements of Figures 4 and 5 is that the filters cannot share storage because they are dispersed. This becomes significant if some of the filters are based on 313-or 625-line delay elements. However, the arrangements of Figures 6 and 7 permit storage sharing.

The relationship between filters F, and F8 and the filters of Figure 1 are given by: F1=Hc-F01F12 F2=-F01 (F2+2 + Fa2-2) H*90 F3= zFF22 [(F3,+' +F3,-') H45 + (F41+1 + F4,-1) H45*] F4=F22+2[F31+1H45+2-F41+1H45*+2]+F22-2[F31-1H45-0F41-1H45*] F5=Hd-F11F02 F6=-F11(F02+2+F02-2)H90* F7 = +F21 [(F321 + F32-')H*45 + (F42+1 + F411)H45] F8 = TF2, [(F32+' + F32-') H45*-(F42+1+F42-1)H45] In these equations H5 and Hd are the transfer functions of the delays of the filters Fo1F,2 and F11Fo2 respectively, and H45 and H90 are the transfer functions of a 45 and 90 phase advance, an asterisk indicating the conjugate (phase retardation). F" represents a transfer function shifted so that its origin lines at nfsc i.e.: Fn(X)=F(x-nfsc) If the filters in the coder path of Figure 1 all use the same delay element i.e. are all based either on 1 or 313 or 625 line delays, then filters F3 and F4 are respectively equal to filters F1 and F2. Similarly if the filters in the decoder path of Figure 1 satisfy the same condition, then filters F7 and F8 are respectively equal to filters F5 and F6. Then the coder and decoder can be further simplified to the arrangements of Figures 8 and 9, where the filters L are luminance-stop filters.In the coder the luminance-stop filter is used to pass the modulated chrominance, thus making the luminance and chrominance spectra complementary. In the decoder the filter L separates out the modulated chrominance for quadrature demodulation.

If the system of Figure 1 uses the one-line-based filters of Figure 2, where Fo1 and F02 are respectively equal to F11 and F12, then the form of L is the same in both coder and decoder and is shown in Figure 10. The decoder compensating delay may be part of L as shown.

The luminance stop filter L shown in Figure 10 will first be described and is seen to consist of two one-line delays 1 50, 1 52 in series, the output of delay 150 constituting an auxiliary output used in the decoder. The average of the input and two-line-delayed signals is taken in a halving adder 154 and this is subtracted from the one-line delayed signal in a subtractor 156. The difference between the input and two-line-delayed signals is also taken in a halving subtractor 158, and the phase of this signal retarded by 90 degrees in a circuit 160. The resultant is multiplied in a multiplier 162 by twice subcarrier at appropriate phase, and the resultant is added to the output of subtractor 156 in an adder 164. The adder output is band pass filtered in a chrominance filter 166.

This filter L is used in both Figures 8 and 9. The filter is seen to provide essentially all the functions of the filters F1 to F4 in Figure 6 and of F5 to F8 in Figure 7. This filter is therefore particularly advantageous in reducing the number of delay elements required.

More generally, the same arrangements are applicable if 313- or 625-line based filters are used throughout, when the delay elements of Figure 10 are modified accordingly. However, in the 313-line case the transfer functions H45 and H90 represent phase shifts of -45 and -90 respectively and the demodulation phase of U is reversed.

Any of the circuit arrangements of Figures 4 to 10 may be realised digitally using an arbitrary sampling frequency in the region of three times the subcarrier frequency. However, a bandpass filter must then precede the modifier wherever it occurs, as shown dotted in Figure 10. The lower cut-off frequency of the bandpass filter must lie between fs-2fsc and 2fs-5fsc, where fas is the digital sampling frequency. The upper cut-off frequency must lie between the nominal video bandwidth and half the sampling frequency.

This application is divided out of application No. 80 05486, Serial No. 2044577.

TABLE Delay basis in lines of luminance and chrominance filters used In Figure 1 and compensating filters

Case F01 F02 F31/F41 F32/F42 1 1D 1D 2n 2n n n 0 0 2 1D 1D 2n+1 2n+1 not n 0 0 3 1 1 2n+1 2n+1 n n O 0 4 2D (group delay n1) 2n2 2n2 n2 > n1 n2-n1 n2-n1 0 0 n2 < n1 0 0 n1-n1 n1-n1 5 20 (group delay n,) 2n,+1 2n2+1 n2 > n1 n2-n1+1 n2-n1 0 0 N2 < n1 1 0 n1-n2 n1-n2 6 10 1D 313 313 313 0 0 0 7 1 t 313 313 312 0 0 0 8 2D (group delay n) 313 313 313 0 n n 9 313 313 313 313 0 0 0 0 10 625 625 313 313 O 0 0 312 0 11 1D 10 625 625 313 312 0 0 12 1 1 625 625 312 312 0 0 13 20 (9roup delay n) 625 625 313-n 312-n 0 0 14 313 313 625 625 0 312 0 0 15 625 625 625 625 0 0 0 0

Claims (3)

1. A PAL coder comprising first, second and third inputs for receiving Y, U and V input signals respectively; V-axis switching means for switching the V signal; multipliers for modulating the U and V signals onto subcarrier at orthogonal phase positions; combining means for combining the U and V signals and subtracting them from the Y input signal; a luminance-stop comb filter connected to the output of the combining means; and a subtractor for subtracting the thus filtered signal from the luminance input signal to provide a PAL output signal.

2. A PAL decoder comprising an input for receiving a PAL input signal; a luminance-stop comb filter connected to the input; a subtractor connected to subtract the output of the luminance-stop filter from the input signal to provide a Y output signal; multipliers for multiplying the output of the luminance-stop filter by two subcarrier signals of orthogonal phase to provide a U output signal and a switched V signal; and V-axis switching means for removing the V-axis switching to provide a V output signal.

3. Apparatus according to claim 1 or 2, in which the luminance-stop filter comprises first and second delays connected in series each of an odd number of lines; half-differencing means for taking half the difference between the input and the output of the second delay; multiplying means for multiplying the output of the half-differencing means buy a signal of twice subcarrier frequency of appropriate phase; and combining means for subtracting the average of the input and the output of the second delay from the output of the first delay and for averaging the resultant and the output of the multiplying means.

3. Apparatus according to claim 1 or 2, in which the luminance-stop filter comprises first and second delays connected in series each of an odd number of lines, combining means for averaging the input and the output of the second delay and subtracting the resultant from the output of the first delay; half-differencing means for taking half the difference between the input and the output of the second delay; multiplying means for multiplying the output of the half-differencing means by a signal of twice subcarrier frequency; and an adder for adding the outputs of the combining means and the multiplying means.

New claims or amendments to claims filed on 25.1 1.82 Superseded claims 1-3 New or amended claims CLAIMS

1. A PAL coder comprising first, second and third inputs for receiving Y, U and V input signals respectively; V-axis switching means for switching the V signal; multipliers for modulating the U and V signals onto subcarrier at orthogonal phase positions; combining means for combining the U and V signals and subtracting them from the Y input signal; a luminance-stop comb filter connected to the output of the combining means; and a subtractor for subtracting the thus filtered signal from the luminance input signal to provide a PAL output signal.

2. A PAL decoder comprising an input for receiving a PAL input signal; a luminance-stop comb filter connected to the input; a subtractor connected to subtract the output of the luminance-stop filter from the input signal to provide a Y output signal; multipliers for multiplying the output of the luminance-stop filter by two subcarrier signals of orthogonal phase to provide a U output signal and a switched V signal; and V-axis switching means for removing the V-axis switching to provide a V output signal.