GB2293073A - Digital ntsc video signals - Google Patents

Abstract

A component of an NTSC colour television signal, e.g. the luminance component, is digitally sampled by sampling an input signal with a mean frequency related to the colour subcarrier frequency, e.g. 2fsc, and changing the sampling phase between successive lines of a field by an amount equal to half the interval between samples. The sampling pattern repeats every four fields, and the sampling phase changes by 180 DEG between corresponding points on adjacent pictures. The phase shift between fields 1 and 2 of the four-field structure can be +90 DEG and the phase shift between fields 3 and 4 can be -90 DEG , or vice versa. This can be used in a coding method which involves sub-band analysis filtering the luminance input video signal into a low-frequency luminance signal and a high-frequency luminance signal, sub-band synthesis filtering the high-frequency luminance signal and the chrominance input video signal to form a combined high-frequency signal, and phase-segregated coding the low-frequency luminance signal and the combined high-frequency signal to provide the composite video output signal. <IMAGE>

Description

DIGITAL NTSC VIDEO SIGNALS This invention relates to the processing of an NTSC colour television signal, and more particularly to the digital sampling of the luminance component of such a signal.

A system for coding PAL colour television signals in a way that minimised loss even when subsequent conversions took place has been described in United Kingdom Patent No. 1,534,268. That method has become known as Weston clean PAL or phase-segregated PAL. The method is described more fully in, for example, EBU Review-Technical, No. 215, February 1986, "A compatible improved PAL system", J.O. Drewery. In the Weston system, the luminance signal is sampled at 2fsc, twice the colour subcarrier frequency, and a line-alternating chrominance signal, consisting of U+V and U-V on alternate lines, is sampled at fsc.

The sampling frequency of 2fsc is a sub-Nyquist sampling frequency, that is it is less than twice the maximum video frequency which is to be transmitted. This sub-Nyquist frequency can only be used because of the line periodicity of the video signal. It can be shown that the main alias components generated in the sampling operation can be removed from the wanted signal by comb filtering.

Such a system could not be readily adapted for NTSC, because in NTSC the alias frequencies would be coincident with the wanted frequencies and thus could not be removed by comb filtering. However, United Kingdom Patent Application No. 2,045,577A (United States Patent 4,322,739) described a method which would allow a Westontype system to be used with NTSC. That application proposed sampling the luminance component by sampling at a mean frequency of twice the colour sub-carrier, but changing the sampling phase between each pair of lines by an amount equal to half the interval between samples (i.e.

1800). This was sufficient to prevent coincidence of the frequency spectra of the wanted and alias signals. This so-called phase-perturbed sampling allows the unwanted signals to be stopped by appropriate comb filters.

The application considered what should happen on a field basis, and in particular considered whether an eight-field periodicity to the perturbed colour subcarrier, or a four-field periodicity, would be better.

The four-field periodicity for the perturbed colour subcarrier, written fsc, would require a 2-field repetitive 2#sc structure for the Y signal. This was thought to have disadvantages and an eight-field periodicity was preferred (see page 7 lines 15 to 26).

We have appreciated that the proposed eightfield periodicity for the phase perturbed colour subcarrier is in fact undesirable, and that the four-field periodicity can be used with advantage, if implemented in a particular way.

First of all, in practice the eight-field periodicity can only be implemented if a special indicator is transmitted with the NTSC signal to indicate the start of the eight-field sequence. This is necessary because there is nothing in the NTSC signal which repeats with a periodicity as long as eight fields, so a special marker is necessary.

Secondly, we have appreciated that an NTSC signal sampled with eight-field periodicity can not readily be used, in the well known 'two-field editing' process.

We have also appreciated that a system with a perturbed sub-carrier implemented with a four-field periodicity in the way we propose actually results in a reduced level of cross-colour arising from static luminance aliases if the signal is decoded in a normal NTSC decoder.

The present invention is defined in the independent claims appended to this description, to which reference should now be made. Advantageous features are set forth in the appendant claims.

In particular, we propose a method of and apparatus for digitally sampling the luminance component of an NTSC colour video signal comprising sampling the input signal with a mean frequency of twice the colour sub-carrier frequency, and changing the sampling phase between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields, and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.

By using a 1800 phase shift between successive pictures, two-field editing is possible. Two-field editing allows a 'field 1' to be made to resemble a 'field 3 so that edits need not be constrained to fit the existing four-field sequence. This is achieved by shifting the picture by half a subcarrier cycle to the left or right, since the NTSC subcarrier changes phase by 1800 over a picture period. For this to work on an NTSC signal according to the present invention, the perturbed subcarrier must therefore also undergo a 1800 phase over a picture period.

Within the four-field structure the 1800 phase shift between fields 1 and 3, and between fields 2 and 4, still leaves flexibility as to the phase shift between -fields 1 and 2, and between fields 3 and 4. We prefer that the phase shift between fields 1 and 2 should be +90 and the phase shift between fields 3 and 4 should be -900.

This has advantages, as described below, in connection with the positions in frequency space of the aliases at the luminance and chrominance band edges in an extended bandwidth system. The plus and minus could be the other way round; that corresponds simply to a redefinition of where the 4-field structure begins.

The invention has especial utility in one particular circumstance. International Patent Application publication NO. W093/22878 describes a method of coding a 625-line component PAL signal into a single signal having a bandwidth of twice the subcarrier frequency, approximately 9MHz, in such a way that the signal closely resembles a PAL signal over the normal PAL signal bandwidth, namely below about 5.5MHz. The system finds particular application in PAL television studios and post-production areas where it is required to produce programmes with a signal quality approaching that available from CCIR Recommendation 601, without the expense of completely re-equipping with component equipment. Use is made of the wider bandwidth available in most studio equipment to convey additional luminance and chrominance detail.

In the coding method described in that application, the luminance is split by sub-band analysis into a low-frequency luminance signal covering approximately frequencies up to the colour subcarrier frequency, and a high-frequency luminance signal comprising frequencies above that. The high-frequency luminance signal is then combined by sub-band synthesis with the chrominance signal to give a combined highfrequency signal. The low-frequency luminance signal and the combined high-frequency signal are then encoded by Weston encoding as though they were the luminance and chrominance signals respectively in the conventional use of the Weston coder. The signal can be disassembled at a special decoder which provides the converse operations.

The present invention can be employed in the Weston encoding of the low-frequency luminance and combined high-frequency chrominance signals in a wide bandwidth component video signal coding method of that type. The resultant coding method can convey a sampled 525-line component signal in a single signal with a bandwidth of approximatefy 7.2MHz, twice the NTSC subcarrier frequency. The coding method forms a single signal from a YW or RGB component video signal in such a way that the spectrum of the signal below approximately 4.6MHz closely resembles an NTSC signal. That is, it can be reproduced on a monitor, for example. The spectrum above this frequency is used to convey additional luminance and chrominance detail.

This use of the invention is described in more detail below, with reference to the drawings. Briefly, the combination of the luminance and chrominance signals around fsc (the NTSC subcarrier frequency) is achieved using phase-segregated or Weston NTSC coding techniques, as described in United Kingdom Patent Application GB 2,045,577. As discussed above, this requires the use of a phase-perturbed colour sub-carrier (referred to as fsc*) and a sampling operation operating at twice this frequency (2fsc*). The structure of the phase-perturbed subcarrier is chosen to have a four-field repeat pattern, contrary to the recommendation of an eight-field pattern in GB 2,045,577, since there is no synchronisation information provided in an NTSC signal to allow a decoding process to be locked to a cycle of length greater than four fields.Furthermore, the structure is chosen to give a phase change of 1800 between frames, in order to allow the use of processes such as two-field editing in which the signal is displaced horizontally by half a subcarrier cycle to convert between the two frame types in the four-field sequence.

Additionally, it is preferred to use a perturbed structure in which there is a phase change of +900 between vertically-aligned points between field 1 and 2, and a change of -900 between such points in fields 3 and 4 of the four-field sequence. This arrangement is beneficial since it causes energy from aliased signals around the luminance and chrominance band edges (caused by the non-sharp-cut nature of the pre-filters and postfilters) to be distributed more evenly in frequency space, rather than being concentrated around particular diagonal frequencies.

BRIEF DESCRIPTION OF THE DRAWINGS The invention will now be described in more detail by way of example with reference to the drawings, in which: Figure 1 is a diagram of the one-dimensional spectrum of the coded signal produced in a preferred embodiment of the invention; Figure 2 is a diagram of the two-dimensional spectrum of the luminance signal transmissible in the preferred embodiment; Figure 3 is a diagram of the two-dimensional spectrum of the chrominance signal transmissible in the preferred embodiment; Figure 4 is a block diagram of the pre-filters and circuitry for forming a combined chrominance signal in the coder; Figure 5 is a block diagram of the circuitry in a decoder to recover two chrominance signals from the combined chrominance signal and to post-filter both luminance and chrominance signals;; Figure 6 is a block diagram of the circuitry in the coder for forming an NTSC-compatible signal from pre-filtered luminance and combined chrominance; and Figure 7 is a block diagram of the circuitry in a decoder for splitting the NTSC coded signal into luminance and combined chrominance.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT In the preferred embodiment, the spectrum of the coded signal is of the form shown in Figure 1.

Chrominance information is shown shaded. The bandwidth of the signal is approximately 7.2MHz (essentially equal to 2fsc, where fsc is 3.5795454545. . MHz), so that it may be sampled without loss at 4fsc, thereby allowing it to be recorded on digital composite video recorders and other equipment operating at this sampling frequency. The spectrum between approximately 4.6MHz and 5.4MHz carries additional chrominance information, essentially in the form of an upper side-band to the normal chrominance signal. The spectrum between approximately 5.4MHz and 7.2MHz carries additional luminance information. The additional chrominance and luminance information does not need to be comb filtered.

Figures 2 and 3 show respectively the twodimensional spectrum of the luminance and chrominance signals transmissible in the preferred embodiment.

Figure 2 shows the way in which the luminance signal is split into low-frequency and high-frequency parts. The figures are drawn for a 16:9 aspect ratio picture.

Samplinq and modulation patterns Before describing the operation of the coder and decoder, the sampling and modulation patterns used will be described. In the following description, the symbol fsc is used to denote the NTSC colour subcarrier frequency, approximately 3.58MHz. The symbol fsc is used to denote the frequency vector which results when the signal frequency of fsc ls combined with the scanning action, and has principal components of 3.58. . MHz, 122 cycles per active picture height. That is to say fsc takes account of the two-dimensional nature of the frequency in horizontal-vertical frequency space, and can be considered by notionally looking at the way the sampling points would be displayed in the image plane.

The most important difference between PAL and NTSC in this application is that the PAL fsc modulation pattern retards by 900 per line whereas the NTSC fsc modulation pattern retards by 1800 per line. In other words, for NTSC, 2fsc is an exact multiple of line frequency and therefore 2fsc is orthogonal, whereas for PAL, 2fsc is approximately an odd multiple of half the line frequency, and therefore 2fsc is quincunxial. In an orthogonal structure, samples on successive lines lie above one another, whereas in a quincunxial structure they are displaced by half a sampling position. Because the assembler and splitter filters that form the heart of the Weston coding system are two-dimensional (i.e. in the horizontal-vertical plane), the sampling operations must be quincunxial to enable the filters to operate as (near) perfect reconstruction sub-band filters.A quincunxial pattern is the only one which can support 2:1 subsampling when the band split is other than purely horizontal or vertical. Therefore, the sampling structure used by the assembler and splitter must have a 1800 phase shift per line.

Therefore it is necessary to perturb the NTSC 2fsc sampling pattern to yield a quincunxial pattern, as described in our aforementioned UK Patent Application GB 2,045,577. This perturbed structure will be referred to as 2fsc*. In this specification fsc* is used instead of #sc as in GB 2,045,577. The perturbation of the 2fsc structure results in a corresponding perturbation of the fsc structure used to modulate the chrominance signal; this perturbed NTSC subcarrier will be referred to as fsct. Since the 2fsc* structure has a 1800 phase shift per line, the fsc* modulation pattern will correspondingly have a 900 phase shift per line, like PAL subcarrier.

The average frequency of the perturbed carrier is that of the NTSC subcarrier, approximately 3.58MHz.

The field-to-field nature of the perturbation may in principle be chosen freely, since it is only the line-to-line nature which determines the correct operation of the assembler and splitter (since they have no memory beyond two lines). However, we have appreciated that there are two criteria that the coded signal should satisfy which limit the choice of perturbed sampling structure: (1). The coded signal should have a four field sequence since there is no synchronisation information in a standard NTSC signal to allow a decoder to be locked to a longer sequence (contrary to what is taught by GB 2,045,577).

(2). The relationship between field 1 and field 3 of the four-field sequence of fsc* should preferably be the same as that between fields 1 and 3 in NTSC fsc (a phase shift of 1800) , in order to allow the use of processes such as two field editing. Such processes convert a 'field 1' to a 'field 3' by shifting the coded signal by half a subcarrier cycle. The same applies to fields 2 and 4.

When these constraints are taken into account, the choice of possible modulation patterns is limited.

There are two choices for the pattern in field 1, these having a line-to-line phase shift of +900 or of -900.

This choice is arbitrary. The pattern in field 2 may be chosen independently to be either of these two patterns; this choice affects the pattern of aliases at the luminance and chrominance band edges and is discussed in more detail below. Additionally, the pattern in field 2 may be positioned in any one of four possible spatial relationships with respect to the pattern in field 1 (samples on adjacent picture lines either aligned vertically or offset by 1, 2 or 3 samples at 4fsc) . The choice of spatial relationship has no material effect on the performance of the system since it merely changes the absolute phase of alias patterns. Once the patterns in fields 1 and 2 are chosen, the patterns in fields 3 and 4 are determined as explained above. Clearly, the same modulation pattern must be used in both coder and decoder.

Therefore in a preferred system the only significant choice that has to be made is whether the modulation patterns in fields 1 and 2 are the same (each either advancing or retarding by 900 per line) or are different (one advancing by 900 whilst the other retards).

The only aspect of the system performance materially affected by this choice is the alias pattern at the horizontal band edges (at frequencies around 3fsc/2 for luminance and fsc/2 for chrominance) , since both the chrominance and high-frequency luminance signals are subsampled on a lattice of fsc*. If the modulation patterns are the same for both fields, alias patterns always appear in the same quadrant of horizontal-vertical frequency space, whereas if they are different the aliases alternate between the first and fourth quadrant on successive fields. Subjectively, the alternating patterns appear to be at a reduced amplitude compared with the fixed patterns. The degree of compatibility with normal NTSC is unaffected by the choice.

We propose to use a pattern in which the modulation phase retards by 90" per line on odd fields but advances by 900 on even fields. We judge the visibility of the alternating alias patterns produced by this arrangement to be slightly lower than that of a nonalternating pattern. This choice means that the performance of the system is identical for lines sloping at a given angle, regardless of whether the slope is to the left or right. It should be noted, however, that the visibility of the alias patterns is generally very low for either case, since they are confined to the luminance and chrominance band edges.

An example of such a perturbed NTSC structure is shown in Table 1 appended to this description. The structures of PAL and NTSC are also shown for comparison.

The numbers indicate the field in the 4-field or 8-field sequence in which the sample falls on a positive peak of the subcarrier. The distance between adjacent columns in Table l represents one sample at 4fsc, and the distance between adjacent rows represents one picture line.

All number 'l's in the perturbed NTSC pattern will be seen to lie in the same pattern as those in the PAL structure.

The pattern in field 2 will be seen to be the mirror image of that on field 1, with samples on a given line displaced by one 4fsc sample to the left of the samples on the preceding line of the same field. The horizontal position of samples in field 2 relative to those in field 1 is arbitrary, as explained earlier. The pattern in field 3 is the same as that in field 1 but horizontally displaced by two samples, giving the required phase shift of 1800 per frame. Field 4 is similarly related to field 2 by a shift of two samples.

One consequence of having fsc* change in phase by 1800 from frame to frame is that 2fsc* is then a framerepetitive pattern. Thus, any alias patterns caused by sampling at 2fsc* appear static, unlike those in the PALcompatible system, which move. The alias pattern is visible for high vertical luminance frequencies around fsc in the picture produced when using a normal NTSC decoder, but is only apparent in the picture from an improved matching decoder when the coder signal has undergone a bandwidth limitation (it appears in the same frequency range).Although a moving pattern might be considered preferable under many circumstances, the static pattern has one advantage when a normal NTSC decoder i-s used, because it gives rise to cross-colour which moves in such a way as to appear less visible than that from either a normal NTSC coder or from a variant of the coding system using a non-frame-repetitive 2fsc* pattern.

Pre-filters and generation of combined chrominance Pre-filters are desirably included in the coder. The luminance pre-filter in the coder, shown in Figure 4 is a horizontal low-pass filter 10 which restricts the bandwidth of the luminance signal to 3fsc/2, i.e. (5.4MHz) so that it can be sampled at a rate of 3fsc.

This is the effective sampling rate used in the subsequent processing. This filtering operation is similar to that for the PAL-compatible system, the difference being that the cutoff frequency is lower, because the NTSC subcarrier frequency is lower than that of PAL. A compensating delay 12 is included to compensate for delays caused by processing in the chrominance channel.

The incoming U and V chrominance signals in the coder are first passed through a matrix 20 to generate signals based on I and Q (the chrominance axes used in NTSC) . In the embodiment shown in Figure 4, the signals generated are Q-I and Q+I, requiring a rotation of -120 from the U and V axes. It is equally possible to implement a coder using a matrix that generates Q and I, in which case minor changes to the vertical modulation process described below are required to achieve functionally equivalent processing. In either case, if the coder already contains a colour matrix for deriving YW from RGB, the two matrix operations may be combined.

The two chrominance signals are then prefiltered vertically by respective filters 22, because the Weston coding method involves vertical subsampling of the chrominance signals by a factor of two. Therefore the two signals are each vertically filtered within a field to remove frequencies in the upper half of the vertical spectrum. The filter is chosen to have a delay of an odd number of half-lines, so that the chrominance signal may be made half a line early with respect to the luminance signal after the compensating delay. This allows for the fact that subsequent circuitry introduces a further oneline delay to the luminance but only half a line delay to the chrominance. This process is common to both the PALand NTSC-compatible coding systems.

The vertically-filtered chrominance signals are then combined to form a single 'combined chrominance' signal by vertical modulation in multipliers 24 and summing in an adder 26. The annotations to the inputs of the multipliers 24 and the output of the adder 26 show the values on four successive field lines in field 1. The values repeat every four lines. The modulation pattern differs from the pattern used in the PAL-compatible system, which was based on the V-axis switch. The pattern should be chosen to generate a combined chrominance signal which, when modulated by fsc* and operated on by the filter F2 in the Weston assembler, yields a modulated signal which resembles a normal NTSC signal sampled on the I and Q axes (the preferred sampling phase for NTSC sampled at 4fsc). This requirement fully determines the form of the switching action.The operation of the Weston assembler is described in the aforementioned patent specification; the main aspect of its performance of importance here is that the filter F2 resembles a (1A, -1A) vertical filter for frequencies around fsc. Table 2 appended to this description shows the switching action required, and how it results in the desired output from F2 (samples are on a lattice at 4fsc).

The Table shows the switched chrominance signal and the perturbed NTSC subcarrier, and illustrates how the combination of these two patterns yields the desired NTSC chrominance modulation pattern after filtering with filter F2. Filter F2 is in the Weston assembler described below with reference to Figure 6. The switching action has a four-line repeat pattern, unlike the PAL case which repeats every two lines. The pattern is different for odd and even fields.

The chrominance switching action is framerepetitive, because in one frame period both fsc* and the required NTSC chrominance output change sign. Therefore the chrominance signal at the input to the fsc* modulator 24 need not change sign from frame to frame. This gives rise to static vertical chrominance aliases when the signal is decoded with a normal NTSC decoder, as distinct from aliases modulated at 12S Hz in the PAL-compatible system. There are corresponding aliases in the picture from a matching decoder, but at such a low level (assuming that the chrominance vertical filters have good stop-band attenuation) that it is largely immaterial whether they are moving or stationary.

It is interesting to note that the spectrum of the switched chrominance signal differs significantly from that in the PAL case. For PAL, U is centred on a carrier at zero vertical frequency and V is on a carrier at 144 c/aph (cycles per active picture height). In contrast, in the NTSC-compatible system, the I and Q signals, in quadrature, are both modulated onto a vertical carrier at odd multiples of 61 c/aph (half of the Nyquist limit for a field, the equivalent of 72 c/aph for a 625-line signal).

Thus they are separated by being in quadrature rather than by being modulated at different vertical frequencies.

Following vertical modulation, the combined chrominance signal is filtered horizontally in a filter 28 to a bandwidth that is supportable by sampling at fsc/2 (1.8MHz), since this is the sampling rate used in subsequent processing. Although this filter is shown as operating on the combined chrominance signal, it could instead operate, for example, on the U and V signals before the matrix, in which case it could be combined with the ADC (analogue to digital converter) pre-filters at the input to the coder.

Post-filters and separation of chrominance Figure 5 shows the circuitry in the decoder corresponding to the circuitry shown in Figure 4. The luminance signal from the sub-band combiner (Figure 7, described below) is applied to a horizontal low-pass filter 30 with cutoff at 3fsc/2, i.e. 5.4MHz, and thence to a compensating delay 32. The combined chrominance signal from the inverse sub-band decoder is applied to a horizontal low-pass filter 40 with cut-off fsc/2, and thence to multipliers 42. Again, the annotations to the inputs of the multipliers 42 show the values on four successive field lines. The values repeat every four lines.

The two signals thus formed, which differ in that the sign of I is reversed, are passed to respective half-band vertical low-pass filters 44 to form Q'+I' and Q'-I', where the primes indicate filtered signals. These are rotated by 120 in a matrix 46 to give U' and V' (the primes indicate signals that have been vertically filtered) The circuits shown in Figure 5 are clearly based on those in Figure 4 and are not therefore described in great detail. It will be seen that the chrominance signals are separated by a vertical quadrature demodulation in multipliers 42 similarly as in the coder, to form two signals of Q-I and Q+I (each vertically subsampled and with the Q and I components in quadrature).

The two vertical low-pass filters 44 remove the repeat spectra caused by the vertical sub-sampling. The filtered signals are then rotated by 120 to produce U' and VR (the rotation may be combined into the YtTV-RGB matrix).

An equivalent implementation is possible in which vertically-subsampled I and Q signals are derived from the combined chrominance signal. After vertical postfiltering, these signals may then be converted to U and V by matrix.

Coder Figure 6 is a block diagram of circuitry embodying the invention for forming a coded signal from the pre-filtered luminance and combined chrominance signals produced by the circuitry of Figure 4.

The pre-filtered luminance signal is applied to an input 100 to which is connected a luminance sub-band splitter 110. The luminance sub-band splitter provides two outputs, the first is a low-frequency luminance signal which is applied to a compensating delay 118, and the second which contains high-frequency luminance. This second signal is applied to a chrominance/high-frequency luminance combiner 120 where it is combined with the prefiltered combined chrominance signal received at an input 102 from the circuit of Figure 4.

The output of the delay 118 and the output of the combiner 120 are applied to the two inputs of a Weston assembler or coder 140. The assembled Weston signal can then be used as desired. As shown, the synchronising and colour burst signals are added at 160 to provide at an output 162 a digital coded output signal sampled at 4fsc.

This can be applied to a digital-to-analogue converter and post-filter 164 to output an analogue coded signal at an output 166.

The composition of the dashed blocks 110,120 and 140 will now be described. The luminance sub-band splitter 110 includes two sub-band analysis filters, namely a low-pass sub-band analysis filter 112 and a highpass sub-band analysis filter 114. The combiner 120 consists of an fsc* sampler 122 coupled to receive the output of the high-pass sub-band analysis filter 114, and an fsc* sampler 124 coupled to receive the pre-filtered combined chrominance signal -Q,-I,+Q,+I, ... received at input 102. The outputs of these samplers are applied respectively to sub-band synthesis filters 126,128. That is, the output of sampler 122 is applied to the input of high-pass sub-band synthesis filter 126, and the output of sampler 124 is applied to the input of low-pass sub-band synthesis filter 128. The outputs of these two filters are then combined in an adder 130.

The structure and operation of the circuits 110,120 is based on the principles described in the abovementioned International Patent Application W093/22878 to which reference should be made for further description thereof.

The Weston assembler 140 comprises a 2fsc* sampler 142 which receives the output of the compensating delay 118 and applies it to a filter 144 with filter function F1. An fsc* modulator 146 receives the output of the combiner 120. The output of modulator 146 is applied to a filter 148 with filter function F2. These filters with functions F1, F2 are inherent to the Weston system, and such filters are discussed in the aforementioned United Kingdom Patent Application 2,045,577. The outputs of the filters 144 and 148 are combined in an adder 150.

It should be noted that the samplers 122, 124 and 142 and the modulator 146 all require a feed of the perturbed sub-carrier signal denoted fsc". This perturbed sub-carrier signal is generated by a modified sub-carrier generator 170. The modifications to a normal sub-carrier generator are not difficult; those skilled in the art who are able to set up a sub-carrier generator for normal operation will be able to set it up for modified operation in the manner described.

The operation of the circuit of Figure 6 will now be considered. The prefiltered luminance signal is split into low-frequency and high-frequency parts using the sub-band filters 112, 114. Sub-band filters are well-known class of filter which allow a sampled signal to be filtered into two or more bands which may then be subsampled and subsequently up-sampled, post-filtered and summed, regenerating the original sampled signal. This is achieved in essence by arranging that the aliases from the subsampling processes exactly cancel losses in the filters.

The sub-band filters used here are twodimensional, splitting the signal as shown in Figure 2.

The reason for using a two-dimensional split is twofold: (1) This shape ensures that the most important luminance frequencies travel in the low-frequency part of the signal: high horizontal frequencies having lower vertical frequencies are generally more important than diagonal frequencies. Thus if the high frequency part of the signal is lost by band-limiting the coded signal, the most important luminance frequencies are retained. This is of particular importance because the high-frequency luminance signal travels frequency-inverted in the coded signal, as discussed below. Therefore a loss of the highest frequencies in the coded signal will cause losses at the lower end of the high-frequency band. If a twodimensional split were not used this would cause the loss of frequencies above fsc for both low and high vertical frequencies.However, with the two-dimensional split of Figure 2, only the higher vertical frequencies around fsc are lost.

(2) The shape of the low-pass filter closely matches the shape of the low-pass filter F1 used in the Weston assembler. Thus the low-pass sub-band analysis filter 112 and the filter 144 in the assembler form a matched pre-filter and post-filter pair, with a 2fsc* sampling operation between them. This ensures that all frequencies passed by the low-frequency sub-band analysis filter travel at the same frequency in the coded signal as they would in a standard NTSC signal. This guarantees good compatibility between the coded signal and standard NTSC.

The output of the low-pass sub-band analysis filter 112 is sampled at 2fsc* (which has a structure that is quincunxial within a field as discussed earlier) by sampler 142. The output of the high-pass sub-band analysis filter 114 is sampled at fsc* by sampler 122, at sites corresponding to positive peaks of subcarrier.

This lower sampling rate is allowed because of the band limitation applied to the signal by the pre-filter.

Furthermore, since the low-pass and high-pass sub-band analysis filters are chosen to have responses very close to zero and unity, respectively, at frequencies beyond 3fsc/2, there is no significant energy from luminance frequencies below fsc/2 travelling as aliases at frequencies in the upper half of the high frequency band which cannot be supported by the lower sampling rate.

The chrominance and high-frequency luminance signals, each sampled at fsc*, are then combined in combiner 120 into a single signal sampled at 2fsc* prior to modulation at fsc* in modulator 146. We have termed the method used to combine these signals inverse sub-band coding. Sub-band coding is usually thought of as a way of allowing a single signal to be split into two signals, each sampled at half the rate of the original, in such a way that the original signal can be recovered exactly.

However, the method can equally be used to combine two independent sampled signals into one signal sampled at twice the rate, in such a way that the two signals may be exactly recovered without interaction or loss. This is achieved by up-sampling each signal by insertion of zero valued samples and filtering using sub-band synthesis filters in exactly the same way as a signal which has been previously split into two sub-bands is regenerated.

The combination of these signals in the coder is therefore achieved by up-sampling the chrominance signal by a factor of 2 to a sampling lattice of 2fsc*, and passing it through a low-pass sub-band synthesis filter 128. The high-frequency luminance signal is similarly processed with the corresponding high-pass filter 126. Note that the up-samplers are not shown explicitly in Figure 6. The filtered signals are added together in an adder 130, creating a signal sampled at 2fsc* with the chrominance signal occupying the lower half of the band and the luminance signal occupying the upper half. There will be some overlap between the two parts of the signal, because the filters will not be infinitely sharp, but this will not prevent perfect separation so long as the filters obey the normal rules for sub-band filters.

Finally, the combined signal, sampled at 2fsc*, is modulated at fsc by modulator 146 as it enters the Weston assembler 140 placing it in the correct part of the spectrum to resemble an NTSC signal.

This processing is essentially the same as that in a PAL-compatible system, except for the use of perturbed sampling structures.

Referring again to Figure 6, the low-frequency luminance signal and the chrominance/high-frequency luminance signals, both sampled at 2fsc*, are then combined in the Weston assembler 140 into a single NTSClike signal sampled at 4fsc. This is accomplished by upsampling each signal by a factor of two to 4fsc by inserting zeroes and using the filters F1 and F2 in a Weston, or phase-segregated, assembler. The up-sampling is not shown explicitly in Figure 6; indeed the arrangement could be implemented with all data paths at 4fsc, so that no rate-changing is required. In such an implementation, the effect of the sampler and modulator is simply to multiply samples by +1,0, or -1 in the appropriate pattern.

The assembler 140 can be considered as a special type of inverse sub-band coder, similar to that used to form the combined chrominance/high-frequency luminance signal described above. As in that application, its task is to form a sampled combined signal from two signals sampled at half the rate in such a way that the two signals can be recovered without loss or interaction. The additional requirement here is that the combined signal should closely resemble a standard NTSC signal. This is achieved by using two-dimensional subband synthesis filters 144,148 to form the combined signal, each filter taking contributions from two successive field lines. Excluding the subcarrier region, however, F1 may be thought of as a low-pass filter, and F2 as a high-pass filter.The most important feature of these filters from the point of view of forming an NTSCcompatible signal is that F2 must behave like a (H,-H) vertical filter at fsc.

The design of the filters in the Weston assembler 140 is discussed in the aforementioned patent specifications and does not need to be described again here.

The signal generated by the assembler is a sampled signal (assumed to be at a rate of 4fsc) containing frequencies in the range O-2fsc. It must be converted into analogue form in such a way that when it is subsequently digitised in a decoder, or other equipment operating at 4fsc, sample values as close as possible to the original 4fsc sample values are obtained. This is important because it is desired to minimise any losses to the signal. If the signal represented a normal NTSC or analogue component signal, a small degree of loss at frequencies near the top of the band would be of little consequence.However, for the signal described here, frequencies near 2fsc carry luminance information from frequencies immediately above and below fsc with high vertical frequencies; as these frequencies are within the passband rather than on the edge it is important to retain them wherever possible.

Therefore, as in GB 2,045,577 the DAC postfilter 164 must be chosen in conjunction with the characteristic of the ADC pre-filter in the decoder to yield a filter product that is approximately Nyquist (skew-symmetric about a response of 1S at 2fsc). This ensures that samples with values very close to the original values are obtained when the signal is re-sampled in the same phase in the ADC. We propose to split the Nyquist filter equally between the DAC and ADC, each filter having a response approximately 3dB down at 2fsc.

Such filters are termed root-Nyquist or half-Nyquist.

An alternative way of achieving the required response without the need for precision analogue filters is to over-sample the signal to, say, 8fsc, and use a digital filter with half-Nyquist response before conversion to analogue form. A similar approach may be used for the ADC.

For interfacing with other equipment operating at a sample rate of 4fsc, such as D2 and D3 recorders, it is possible to use the digital output 162 of the coder directly.

Decoder The corresponding decoder is shown in Figure 7.

This is the precise inverse of the circuit of Figure 6, and it is not necessary to provide a detailed textual description of the figure in view of the above detailed description of Figure 6. Corresponding elements have corresponding reference numerals increased by 100. The only additional component is an adder 216 in the luminance sub-band combiner 210.

Briefly, in the decoder shown in Figure 7, the splitter filters F3 and F4 act as sub-band analysis filters that match the synthesis filters F1 and F2. The output of F3 is sampled at 2fsc* in sampler 242 to yield nominally identical sample values to those entering F1.

The output of F4 is modulated at fsc* (a process which is equivalent to sampling at 2fsc* and inverting alternate samples) in modulator 246; this yields sample values nominally identical to those entering the modulator prior to F2. The high-frequency luminance and chrominance signals are separated by the use of matching sub-band analysis filters 226, 228. The filtered signals are resampled at fsc* in samplers 222, 224 to yield sample values identical to the original signals in the coder.

Finally, the luminance signal is reconstructed from the sampled low-frequency and high-frequency subbands using sub-band synthesis filters 212, 214. Since the high-frequency sub-band has been sub-sampled, there will be an aliased signal present in the upper half of the high-frequency sub-band, which will appear as an alias in the upper quarter of the reconstructed signal at frequencies between 3fsc/2 and 2fsc. This alias is removed by the luminance horizontal post-filter shown in Figure 5, which was described above.

Characteristics of the NTSC-compatible system The most significant difference in performance of the NTSC-compatible system compared with the PALcompatible system is the reduced horizontal bandwidths of luminance and chrominance. This is a direct consequence of the lower subcarrier frequency used, and hence the lower overall bandwidth available in equipment such as D2 -and D3 recorders operating at 4fsc. The luminance bandwidth, 3fsc/2, is approximately 5.4MHz; the chrominance horizontal bandwidth is a third of this figure, approximately 1.8MHz. Thus the NTSC-compatible system falls a little short of allowing the full luminance bandwidth of CCIR Recommendation 601 to be conveyed.

The ratio of horizontal to vertical resolution of the NTSC-compatible system is close to the corresponding ratio for the PAL-compatible system, since the ratio of the subcarrier frequencies (approximately 0.81) is similar to the ratio of the number of active lines (approximately 0.85). Therefore the response of the system remains well-matched to a 16:9 display.

The fact that the subcarrier frequency is lower for NTSC compared with PAL allows a wider combing region =o be used in the assembler and splitter filters, whilst still limiting the upper end of the region to a frequency (around 5.5MHz) likely to be passed by all types of studio equipment (limiting the combing region to frequencies likely to always be passed ensures that no cross-effects will ever be present). The wider combing region improves the compatibility of the signal with normal NTSC because the chrominance signal more closely resembles the doublesideband signal of NTSC.One consequence of this is that since the amplitude of the chrominance signal remains constant for a wider range of frequencies around fsc, the appearance of slight luminance-chrominance misregistration when a normal NTSC decoder is used (discussed for the PAL case in PCT/GB93/00870) is significantly reduced, because filter F2 resembles a ('A -'A) vertical filter over a wider range of frequencies. This is particularly advantageous, because it is not possible to compensate for any mis-registration by applying a fixed relative delay to one chrominance component (as can be done in the PAL-compatible system) because of the different nature of the vertical chrominance modulation. TABLE 1 PAL NTSC Perturbed NTSC (fsc*) 1 3 5 7 1 3 5 7 1 . 3 . 1 . 3 . 1 3 . 1 . 3 .

4 6 8 2 4 6 8 2 4 . 2 . 4 . 2 . 2 4 . 2 . 4 .

7 1 3 5 7 1 3 5 3 1 . 3 . 1 . . 1 3 . 1 . 3 2 4 6 8 2 4 6 8 2 . 4 2 . 4 4 2 . 4 . 2 5 7 1 3 5 7 1 3 1 . 3 . 1 3 . 3 . 1 . 3 . 1 .

8 2 4 6 8 2 4 6 4 . 2 . 4 . 2 . 4 . 2 . 4 2 3 5 7 1 3 5 7 1 3 . 1 . 3 1 . . 3 . 1 . 3 1 6 8 2 4 6 8 2 4 2 . 4 . 2 . 4 . . 2 4 . 2 . 4 TABLE 2 Field line (chrominance signal fsc* modulation Signal at input to F2 Output of F2 from switch) (ignoring scaling factor) 1 -Q 0 + 0 - 0 -Q 0 +Q +I +Q -I -Q 2 -I - 0 + 0 +I 0 -I 0 -I -Q +I +Q 3 +Q 0 - 0 + O -Q O +Q +I +Q -I -Q 4 +I + 0 - 0 +I 0 -I 0 -I -Q +I +Q Field 2 1 +Q 0 + 0 - 0 +Q 0 -Q -I -Q +I +Q 2 -I + 0 - 0 -I 0 +I 0 +I +Q -I -Q 3 -Q 0 - 0 + 0 +Q 0 -Q -I -Q +I +Q 4 +I - 0 + 0 -I 0 +I 0 +I +Q -I -Q

Claims (10)

1. CLAIMS 1. A method of digitally sampling a component of an NTSC colour television signal, comprising sampling an input signal with a mean frequency related to the colour subcarrier frequency, and changing the sampling phase between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields, and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.
2. 2. A method according to claim 1, in which the phase shift between fields 1 and 2 of the four-field structure is +90 and the phase shift between fields 3 and 4 is -900, or vice versa.
3. 3. A method according to claim 1, in which the input signal is a luminance signal and is sampled at a mean frequency of twice the colour sub-carrier frequency.
4. 4. Apparatus for digitally sampling a component of an NTSC colour television signal, comprising sampling means for sampling with a mean frequency related to the colour subcarrier frequency, and phase perturbing means for changing the sampling phase between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields, and in which the phase perturbing means causes the sampling phase to changetby 1800 between corresponding points on adjacent pictures.
5. 5. Apparatus according to claim 4, in which the phase shift between fields 1 and 2 of the four-field structure is +900 and the phase shift between fields 3 and 4 is -900, or vice versa.
6. 6. A signal at NTSC colour subcarrier frequency, in which the signal has a four-field repetitive structure, and the phase changes by +90 between successive field lines and by 1800 between successive pictures.
7. 7. A method of coding luminance and chrominance input video signals to provide a composite video output signal which consists of: (a) a signal representing low frequency luminance information predominantly occupying in the composite signal frequencies in the range from zero to the colour subcarrier frequency; (b) a signal representing chrominance information predominantly occupying in the composite signal frequencies from colour subcarrier frequency to a predetermined higher frequency; and (c) a signal representing high luminance frequencies predominantly occupying in the composite signal frequencies from the predetermined higher frequency to an upper limit; the method comprising the steps of: sub-band analysis filtering the luminance input video signal into a low-frequency luminance signal and a high-frequency luminance signal; sub-band synthesis filtering the highfrequency luminance signal and the chrominance input video signal to form a combined high-frequency signal; and phase-segregated coding the low-frequency luminance signal and the combined high-frequency signal to provide the composite video output signal; in which the phase-segregated coding includes sampling the low-frequency luminance signal at a mean frequency of twice the colour subcarrier frequency and modulating the combined high-frequency signal at a mean frequency equal to the colour subcarrier frequency, the sampling phase changing between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields; and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.
8. 8. Apparatus for coding luminance and chrominance input video signals to provide a composite video output signal which consists of: (a) a signal representing low frequency luminance information predominantly occupying in the composite signal frequencies in the range from zero to the colour subcarrier frequency; (b) a signal representing chrominance information predominantly occupying in the composite signal frequencies from colour subcarrier frequency to a predetermined higher frequency; and (c) a signal representing high luminance frequencies predominantly occupying in the composite signal frequencies from the predetermined higher frequency to an upper limit; the apparatus comprising: luminance signal input means fdr receiving a luminance input video signal; chrominance signal input means for receiving a chrominance input video signal; sub-band analysis filter means coupled to the luminance signal input means for filtering the luminance input video si'gnal to provide at a first output a low-frequency luminance signal and at a second output a high- frequency chrominance signal; sub-band synthesis filter means coupled to the chrominance signal input means and to the second output of the sub-band analysis filter means for filtering the chrominance input video signal and the high-frequency luminance signal to provide at its output a combined highfrequency signal; and phase-segregated coding means coupled to the first output of the sub-band analysis filter means and to the output of the sub-band synthesis filter means for coding the low-frequency luminance signal and the combined high-frequency signal to provide a composite video output signal; in which the phase-segregated coding means includes means for sampling the low-frequency luminance signal at a mean frequency of twice the colour subcarrier frequency and modulating the combined high-frequency signal at a mean frequency equal to the colour subcarrier frequency, the sampling phase changing between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields; and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.
9. 9. A method of decoding a composite video input signal which consists of: (a) a signal representing low frequency luminance information predominantly occupying in the composite signal frequencies in the range from zero to the colour subcarrier frequency; (b) a signal representing chrominance information predominantly occupying in the composite signal frequencies from colour subcarrier frequency to a predetermined higher frequency; and (c) a signal representing high luminance frequencies predominantly occupying in the composite signal frequencies from the predetermfned higher frequency to an upper limit; to provide luminance and chrominance output signals, the method comprising the steps of: phase-segregated decoding the composite video input signal to provide a low-frequency luminance signal and a combined high-frequency signal; sub-band analysis filtering the combined high-frequency signal to provide a high-frequency luminance signal and a chrominance output signal; and sub-band synthesis filtering the lowfrequency luminance signal and the high-frequency luminance signal to provide a full-bandwidth luminance output signal; in which the phase-segregated decoding includes sampling the low-frequency luminance signal at a mean frequency of twice the colour subcarrier frequency and modulating the combined high-frequency signal at a mean frequency equal to the colour subcarrier frequency, the sampling phase changing between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields; and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.
10. 10. Apparatus for decoding a composite video input signal which consists of: (a) a signal representing low frequency luminance information predominantly occupying in the composite signal frequencies in the range from zero to the colour subcarrier frequency; (b) a signal representing chrominance information predominantly occupying in the composite signal frequencies from colour subcarrier frequency to a predetermined higher frequency; and (c) a signal representing high luminance frequencies predominantly occupying in the composite signal frequencies from the predetermtned higher frequency to an upper limit; to provide luminance and chrominance output signals, the apparatus comprising: input means for receiving a composite video input signal; phase-segregated decoding means coupled to the input means to provide at a first output a lowfrequency luminance signal and at a second output a combined high-frequency signal; sub-band analysis filter means coupled to the second output of the phase-segregated decoding means to provide at a first output a high-frequency luminance signal and at a second output a chrominance output signal; and sub-band synthesis filter means coupled to the first output of the phase-segregated decoding means and to the first output of the sub-band analysis filter means to provide a full-bandwidth luminance output signal; and in which the phase-segregated decoding means includes means for sampling the low-frequency luminance signal at a mean frequency of twice the colour subcarrier frequency and modulating the combined highfrequency signal at a mean frequency equal to the colour subcarrier frequency, the sampling phase changing between successive lines of a field by an amount equal to half the interval between samples, the sampling pattern repeating every four fields; and in which the sampling phase changes by 1800 between corresponding points on adjacent pictures.