GB2051517A - Digital television transmission system and encoder and decoder therefor - Google Patents

Abstract

A digital transmission system for PAL (or NTSC) colour television signals encodes the samples so that successive ones of the samples represent respective ones of the luminance and chrominance components, with the position of each individual sample in the stream, whether luminance or chrominance, being related to the position on the television raster of the signal element which only that sample describes. Furthermore, the luminance samples are taken at the sample rate of the encoded signal with selected samples, e.g. every one in three, being omitted. The omitted sample periods are occupied in the encoded signal by chrominance signals. To this end in a coder (10), Y, U and V component signals are digitised (32, 34, 36) and the U and V signals multiplexed (40). The resultant chrominance signal is multiplexed (42) into the luminance signal by replacing every third luminance sample by a chrominance sample. At the decoder (12), the samples are demultiplexed (46, 48) and the missing luminance samples estimated (50) from the transmitted ones. <IMAGE>

Description

SPECIFICATION Digital television transmission system and encoder and decoder therefor This invention is concerned with the transmission or processing of digital colour television signals, such as PAL or NTSC signals, with the signals in their component form, i.e. separated into luminance and un-modulated chrominance signals. It is anticipated that the normal use of the invention will be in transmission and this term will hereinafter be used to include other forms of processing.

The following description will be given principally in terms of the 625/50 PAL System I colour television system in use in the United Kingdom. It is not, however, limited to such application.

To transmit a PAL television signal, a serial digital sample stream could be formed from regularly-sampled luminance and colour difference signals by time-division multiplexing the luminance and colour separation signals. Such an arrangement might lead to a sample train as follows: Sample No. : 1 2 3 4 5 1 2 3 4 5 1 Sample : Y Y U V Y Y Y U V Y Y In this sequence, the U samples occur regularly at position 3 and the V samples at position 4.

The Y samples, which are regularly taken in time from the original Y signal, occur at positions 1, 2 and 5. The U and V samples at positions 3 and 4 form a group with the luminance sample at position 2, in that the three samples of the group relate to identically the same point on the television raster.

This sample stream would be difficult to process in some applications, because the positions of the samples in the sample stream would not relate directly to the positions of the signal elements they describe on the television raster. Instead, the sample stream would contain groups of samples (2, 3 and 4) which would need to be treated similarly. This might affect such processes as mixing and switching between television signals, where the switching waveform position would have to be related to the sample groups.

There is a proposal to standardise on a sample rate of approximately four times the colour subcarrier frequency (4 fsc) for the digital television transmission of composite signals. There is thus a problem in devising an efficient sampling scheme for the separated components of a colour television signal which would be compatible with such a 4fisc standard for composite signals. It is undesirable to use a sample rate of only 2fisc for the luminance signal, as this is a sub-Nyquist frequency and some aliasing or spectrum folding is inevitable.

The present invention seeks to provide some reduction in one or other of the foregoing problems, and is defined in its various aspects in the appended claims, to which reference should now be made.

The invention will be described in more detail, by way of example, with reference to the drawings, in which: Figure 1 is a block circuit diagram of a digital colour television transmission system embodying the invention for use with PAL signals and comprising a coder and decoder; Figure 2 illustrates in principle a sample structure where one in three luminance samples are replaced by a chrominance sample, with the sequence reset on each line; Figure 3 illustrates in similar form a simple YUV sample structure; Figure 4 illustrates in similar form an alternative YUV sample structure.

Figure 5 is a block circuit diagram of a comb filter which can be included in the chrominance signal paths at the encoder; Figure 6 is a block circuit diagram of a corresponding comb filter which can be used at the decoder; Figure 7 is a block circuit diagram of a circuit for use at the coder to multiplex the content of the chrominance signals into U + V alternate sample form; Figure 8 illustrates the sample structure obtained using the circuit of Fig. 7; Figure 9 is a block circuit diagram of a spatial filter for use at the decoder to separate U and V from the sample structure of Fig. 8; Figure 10 illustrates the signal response of the filter of Fig. 9 in the U signal path; Figure ii is a block circuit diagram of a system similar to Fig 1 for use with NTSC-type Y, I and Q signals;; Figure 12 shows the sample structure obtained with the coder of Fig. 1 1; and Figure 13 illustrates a possible rearrangement which can be used with the coders of Figs. 1 or 11.

Fig. 1 shows a coder 10 and a decoder 1 2 separated by a diagrammatically illustrated transmission channel 14. The coder 10 has three inputs 16, 1 8 and 20 for receiving respectively analogue Y, U and V signals and has a single output 22 for providing an encoded output sample stream. The decoder 12, conversely, has a single input 24 and three outputs 26, 28 and 30 for delivering Y, U and V signals.

At the coder 10, the three input signals Y, U and V are sampled by respective analogue-todigital converters 32, 34 and 36 operating at a common sampling frequency f, in accordance with pulses received at an input 38. The chrominance signals U and V are combined by a multiplexer 40 into a single sample stream having a sample rate of f,/n, where n is an integer and is typically 3 or 4. It will be appreciated that the ADCs 34 and 36 each need in fact only operate at a sampling frequency of fs/2fl The outputs of the multiplexer 40 and the ADC 32 in the luminance path are applied to respective inputs of a selector switch 42 which is controlled in accordance with a control switching waveform received at an input 44.The control waveform instructs the switch 42 to select the output of multiplexer 40 on one in n of the sample periods and on the other sample periods to select the luminance output of the ADC 32.

This has two consequences. Firstly, considering the luminance path, one in n of the luminance samples is discarded, and is replaced by a colour difference sample. Secondly, in the resultant sample stream, the positions of the samples in the sample stream relate directly to the positions on the television raster of the signal elements they describe, or from which they are derived. In other words, the colour samples do not relate to any of the transmitted luminance samples; in fact they relate to the omitted luminance samples.

The sample frequency f, is preferably equal to four times the colour subcarrier frequency fsc, or to a nearby integral multiple of the line frequency. The sample rate of the encoded signal at the output 22 of the coder 10 is equal to the initial sample rate f, of the luminance signal. This system makes efficient use of the sample rate available.

At the decoder 12, the colour difference samples are separated from the luminance samples by a switch 46 which is controlled in synchronism with the switch 42 at the coder. The colour difference samples are applied to a demultiplexer 48 which supplies U and V samples to the outputs 28 and 30 respectively.

The luminance samples from switch 46 are applied to an estimating filter 50 and a compensating delay 52, the outputs of which are applied to a further selector switch 54. This switch is controlled in synchronism with the other switches 42, 46 subject only to a delay equal to the compensating delay 52, and applies the Y samples to the output 26. The estimating filter 50 estimates the missing one-in-n luminance samples from the values of the luminance samples which are transmitted.

The estimating filter 50 can be of similar construction to the digital transversal filter shown in Fig. 2(a) of BBC Research Department Report 1977/27 (C. K.P. Clarke). The estimating filter 50 accurately determines missing sample from surrounding sample values on the basis that the digital video signal contains signal energy up to 5.5 MHz only. The estimating filter incorporates a transversal measurement filter as described in the above mentioned Report, the output of the measurment filter being subtracted from the input video signal. The timing relationship in the subtraction is made to be such that the temporal location of the missing sample is put at the central position of the transversal measurement filter. Thus, any data at that location then plays no part in estimating the value of the missing sample.Hence, an estimating filter is derived from a measurement filter by negating each term and setting the central term to zero.

The filter parameters will now be considered. To estimate every fourth sample of a 4fisc sampled signal, for example, it is a requirement of the time response of the estimating filter that every fourth coefficient shall be zero. Also, to be a good estimating filter, it should have unity gain within the 5.5 MHz video passband.

To assess whether such a filter is feasible, we have designed some, and computed their responses. A first filter has 1 5 coefficients symmetrically either side of the point being estimated. The coefficients are given by: n sin null/4 an = (1 +COS nII/20) null/4 this being a combination of the functions (sin x)/x and raised cosine. Note that for n = O or any multiple of four, then an = 0.

Two alternative filters had the number of coefficients reduced (16 being substituted for 20 in the rasied cosine part of the expression given above). The two alternatives give different ways of quantising the coefficients to nine bit accuracy. The coefficients are given in the following table: Coefficients (an) a1 a2 a3 a4 a5 a6 a7 a9 a10 a11 a13 a14 a15 First alternative + 228 - 156 + 70 - 35 + 37 - 19 + 10 - 10 + 4 - 1 + 0 - 0 Second alternative + 228 - 157 + 70 - 36 + 38 - 20 + 10 - 10 + 5 - 1 + 1 - 0 In this table, the coefficents are given multiplied by 256. The coefficents aO, a4, a8, a,2 and a16 are omitted as they are all zero.

Regarding the estimating filter 50, reference should be made to our British Patent Appiication No. 49293/78.

The control signals for the switch 42 at the coder 10 can be derived from the television signal. They can then either be multiplexed into the serial sample stream, and thus conveyed to the decoder using a low data rate, or once again be derived from the television signal at the decoder 12.

A typical sample structure obtained using the system of Fig. 1 is illustrated in Fig. 2, where n is 3 and the phase of sampling of the colour difference signals has been reset during each lineblanking interval, so that the sample pattern is the same on each line. Here the chrominance samples are of unspecified content and are indicated simply by zeroes.

When applied to the YUV component signals of a PAL television system, conveniently either one in three or one in four luminance samples could be replaced by chrominance samples.

However, replacing one in three would appear to be a better choice, as it leads to a more appropriate subdivision of the sample rate between luminance and chrominance components.

A number of examples will now be given of ways in which the U and V colour difference signals could be processed, bearing in mind that the available sample rate of one in three samples of a 4fisc signal, is only 4f,,/3 (i.e. 4/3 X 4.43 MHz = 5.91 MHz). In each case the luminance processing would be as shown in Fig. 1.

Regular sampling of the U and V colour difference signals could be done using the sample structure shown in Fig. 3. This structure has the advantage of being repeated on each line.

However, with a sample rate of only 2.96 MHz for each of the U and V signals, these signals would have to be band restricted by a greater degree than would be normal for colour difference signals intended for a PAL signal.

A better sample structure might be as shown in Fig 4. Because the sampling phase of each colour difference signal is changed by 180 on alternate lines, the band restriction of these signals before sampling could be less severe. Some aliasing could be permitted, particularly bearing in mind that the most common form of PAL decoder, to which this signal would eventually be applied after encoding into a composite PAL signal, contains a line-delay comb filter.

In order to enable a sub-Nyquist sampling technique to be applied to the U and V signals to give them a greater effective horizontal bandwidth, the high-frequency components of the U and V signals may be line-delay comb filtered between, say, 1 MHz and 2 MHz. Suitable apparatus for use in the U signal path of the coder 10 is shown in Fig. 5 and apparatus for use at the decoder 1 2 is shown in Fig. 6.

Referring to Fig. 5, a demultiplexing switch 60 distributes incoming samples alternately to a line delay 62 and to the inverting input of a halving subtractor 64. The non-inverting input of the subtractor receives the output of the one-line delay. The subtractor output is applied to a high-pass filter 66 with a cut-off frequency of about 1 MHz. Finally, a subtractor 68 subtracts the output of the filter 66 from the output of delay 62, to provide the output of the comb filter which is then applied to the multiplexer 40 of Fig. 1.

In Fig. 6, the U signal from the demultiplexer 48 in the decoder 1 2 is applied to the noninverting input of a subtractor 70 and through a 1 MHz high-pass filter 72, which also operates as a halving circuit, to the inverting input of the subtractor 70. The filter output is also applied to a one-line delay 74. A multiplexing switch 76 alternately selects samples from the subtractor 70 and the delay 74 and applies them to the U output 28.

When the comb filters of Figs. 5 and 6 are connected in series, as they are in the transmission system as a whole, an average of one line delay is introduced in both the high and Icw frequency components of the U (or V) signal. Consequently, a one-line delay should also be included somewhere in the luminance path, either before switch 42 in the coder 10 or, preferably, after switch 46 in the decoder 1 2.

The filters of Figs. 5 and 6 are similar to those described by V. G. Devereux and A. H. Jones in " Pulse code modulation of video signals: codes specifically designed for PAL", Proc. IEE, Vol.

125, No. 6, June 1978, pages 591 to 598. Fig. 9 of that paper shows apparatus for 2fisc sub Nyquist sampling of PAL signals. However, in the present case the initial sampling frequency of the U and V components is 5.91 MHz, and the final sampling frequency is 2.96 MHz. The sample structure for this system would be as shown in Fig. 4.

The particular benefits obtained with this system are that a wide (e.g. 2 MHz) bandwidth is achieved for each of the U and V components, and any impairments caused by loss of high spatial frequencies would not noticeably affect the quality of the picture displayed on a domestic receiver, where the chrominance channels are typically fairly narrow band.

As decribed above, the colour samples have been sent in the form of U and V samples alternately in time. It may, however, be preferable to multiplex them in content, particularly if a spatial filter is used at the decoder.

Fig. 7 illustrates a suitable form for the multiplexer 40 to achieve this form of multiplexing.

The V signal from ADC 36 operating at 5.91 MHz is applied to an inverter 80. A switch 82 alternately selects the inverted signal from inverter 80 and the non-inverted signal from ADC 36. The resultant signal consists of + V and - V on alternate colour samples, and is added to the U signal in an adder 84. The resultant multiplexed signal consists of U + V and U - V on alternate colour samples, as shown in Fig. 8.

At the decoder 12, a spatial filter 100 of the form shown in Fig. 9 could be used to separate the U and V samples. The filter has an input 102 which is connected to the output of switch 46 and to which is connected a series of delays 104, 106, 108 and 11 0. The length of delays 106 and 108 is one sample period T of the colour signal, i.e. one period at a frequency of 5.91 MHz, while the length of delays 104 and 110 is equal to one line period (H) less one such sample period (T). The output of delay 106 is multiplied by a factor a in a multiplier 11 2 and applied to one input of an adder 114. The input signal and the outputs of delays 104, 108 and 110 are summed in a summing circuit 1 16.The output of the summing circuit 116 is multiplied by a factor b in a multiplier 1 18 and applied to the other input of the adder 114.

From inspection of Fig. 8, it will be seen that at any instant the samples at the junctions marked A, B, D, and E in Fig. 9 will all be of one type (e.g. U + V), while the sample at junction C will be of the other type (U - V). To obtain a U signal output, coefficient a is T and coefficient b is 18. For the V signal output, coefficient a is again + and coefficent b is - Xf; the output 1 20 is then applied to a circuit (not shown) which multiplies each alternate output sample by - 1.

Such a system allows a particularly wide bandwidth (e.g. 2.5 MHz) to be achieved for each of the U and V colour difference signals at some loss of vertical resolution.

The horizontal and vertical filtering actions of a decoder including the circuits of Figs. 7 and 9 are given in Fig. 10 for the U signal, and also apply analogously for the demodulated V signal. From this Fig., it can be seen that the losses of both vertical and horizontal resolution would be minimal and would probably not affect the displayed signal quality. However, as is also shown in Fig. 10, the high spatial frequency V components are aliased into the U channel and vice versa. Such aliasing effects would probably be acceptable and would in any case be reduced by the filtering action of the line-delay comb filter in the final PAL decoder after the signal has been converted into its composite PAL form.

Fig. 11 illustrates a system designed for use with signals in YIQ form intended for eventual modulation to form a composite NTSC colour television signal. Corresponding components to those of Fig. 1 are given the same references, the I and Q signals being taken to correspond to the U and V signals respectively.

In Fig. 11 the multiplexer 40 for multiplexing the I and 0 signals is shown as a selector switch 40. The ADC 34 samples the I colour difference signal at 4fsc/3. One in every four of the I samples is then replaced by a Q sample from ADC 36, which consequently samples at a rate of fsc/3. Thus, the sample omission system is applied twice. Switch 40 is controlled from an input 1 22. It will be appreciated that the switches 40 and 42 could be combined into a single selector unit.

With this system, and assuming a colour subcarrier frequency of 3.58 MHz, the mean sample rates for the Y, I and Q signals are: For Y; 8fisc/3 = 9.54 MHz For I; (4) X (4fsc/3) = 3.58 MHz For O; (-) x (4fsc/3) = 1.19 MHz.

The total chrominance sample rate is 4.77 MHz.

A suitable sample structure is shown in Fig. 1 2. The structure shown assumes that the sample phae of both I and Q signals is reset each television line.

As seen in Fig. 11, the omitted I samples are estimated from the transmitted I samples with the aid of an estimating filter 1 24 provided with a selector switch 1 26 and compensating delay 128, which are analogous to the corresponding components 50, 54 and 52 in the luminance path.

With the systems shown in Figs. 1 and 11, no two samples which are transmitted through the transmission channel 14 are sited at precisely the same points on the television signal raster.

Thus, the three analogue-to-digital converters 32, 34 and 36 can be replaced by a single converter 1 30 connected after the switch or multiplexer 42, as shown in Fig. 1 3. All processing up to the ADC 1 30 is done with analogue signals. In this way, the need for three (relatively expensive) ADCs is avoided.

Claims (11)

1. A method of digitally coding a colour television signal in luminance/chrominance component form, in which successive samples represent respective ones of the luminance and chrominance components, and the position of each individual sample in the stream is related to the position on the television raster of the signal element which that sample describes.

2. A method according to claim 1, in which the luminance samples are taken at the sample rate of the encoded signal with selected samples being omitted, and the omitted sample periods are occupied in the encoded signal by chrominance samples.

3. A method of digitally coding a colour television signal in luminance/chrominance form, in which luminance samples ar taken at the sample rate of the encoded signal with selected samples being omitted, and the omitted sample periods are occupied in the encoded signal by chrominance sampleS.

4. A method according to any preceding claim, in which the proportion of chrominance samples is from one in three to one in four of the samples in the encoded signal.

5. A method according to any preceding claim, in which the chrominance signals are linedelay comb filtered before encoding.

6. A method according to any of claims 1 to 5, in which the chrominance samples in the encoded signal each relate to one of the component chrominance signals of the colour television system.

7. A method according to any of claims 1 to 5 in which the chrominance samples in the encoded signal relate to at least two different combinations of the component chrominance signals of the colour television system.

8. A method according to any preceding claim, in which the component signals are multiplexed in analogue form and are then digitised to provide the encoded signal.

9. A method of digitally coding a colour television signal substantially as herein described with reference to the drawings.

1 0. A coder for use in the method of claim 1, comprising inputs for receiving television signals in luminance/chrominance component form, and multiplexing and sampling means connected to the inputs for producing an encoded signal successive samples of which represent respective ones of the luminance and chrominance components, with the position of each individual sample in the encoded signal being related to the position on the television raster of the signal element which that sample describes.

11. A coder for use in the method of claim 3, comprising inputs for receiving television signals in luminance/chrominance component form, and multiplexing and sampling means connected to the inputs for producing an encoded signal, in which luminance samples are taken at the sample rate of the encoded signal with selected samples being omitted, and the omitted sample periods are occupied in the encoded signal by chrominance samples.

1 2. A decoder for use with the coder of claim 11, comprising demultiplexing means for separating the encoded samples into luminance and chrominance samples, and means connected to receive the luminance samples and adapted to estimate values for the omitted samples from the luminance samples in the encoded signal.

1 3. A coder for digital colour television signals, substantially as herein described with reference to the drawings.

1 4. A decoder for digital colour television signals, substantially as herein described with reference to the drawings.