GB2106348A - Video signal coding - Google Patents

Abstract

A video signal is encoded by a recursive coding procedure such as differential pulse code modulation (DPCM). Individual lines of the video signal are distributed on a line-by-line basis cyclically between two (or more) processing channels and re-timed thereby reducing the signal rate in each channel. Each channel has its own encoder (11-16; 21-26) whose outputs are combined (4) to produce the desired encoded output. The encoders can be interconnected (16, 26) so that at least part of the preceding line is available for the recursive coding procedure. <IMAGE>

Description

SPECIFICATION Video signal coding This invention relates to a method of an apparatus for transmitting or processing a video signal, and which makes use of a recursive encoding system such as for example differential pulse code modulation (DPCM).

DPCM is a well known technique by which digital video information can be transmitted in differential form: a "guess" or prediction of a sample is made and the difference between the actual and predicted sample is transmitted. If the prediction is good, this difference is small and a reduction in transmitted bit rate can ensue. The structure of a DPCM coder is inherently "recursive" in the sense that the decoded Kth value must be available for the (K+1 )th sample to be worked out. This limits the overall speed at which the system can work with a given circuit technology as it would appear that conventional "pipe-lining" techniques cannot be used.

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

One embodiment of the invention will be described by way of example with reference to the accompanying drawings in which: Fig. 1 is a block circuit diagram of an encoder embodying the invention.

Fig. 2 is a timing diagram illustrating the store operation of the apparatus of Fig. 1; Fig. 3 is a further timing diagram illustrating the relative positions of samples in successive television lines, and Fig. 4 shows in more detail the weighting arrangements of the apparatus of Fig. 1.

In the apparatus shown in Fig. 1, an input video signal is fed to an electronic switch X1 which is controlled to direct alternate lines of the video waveforms to respective line stores 10, 20.

Successive samples of the video waveform are entered into the stores in real time, by means of a source 2 of clock pulses synchronous with the video sampling rate, but read out at half this rate, read-write control circuitry (not shown) being provided to control the store operation and to switch over to half-rate clock pulses from a divider 3 during readout. The store timing is illustrated in Fig. 2.

The output from each store is fed to a respective DPCM coder arrangement. The video signal from each line store 10,20 is taken to a subtractor 1 1,21 where the difference between the current sample and a predicted value is taken and fed via a quantiser 12,22. The quantiser outputs are fed to as combining unit 4 which may for example convert the two digital quantised signals into serial form, interleaved for transmission. The quantised signals are added (in adders 13,23) to the respective predicted signals to form an output which corresponds to the outputs which will be obtained in the decoder at the receiving end.The predicted signal is formed in predictors 14,24 taking into account the values of one or more samples preceding the sample which is to be predicted, samples taken two lines earlier, via a delay 1 5,25-which is actually the preceding line in the channel in question, due to the separation of alternate lines into separate channels-and samples from the preceding line, which are obtained via half-line delays 16,26 from the adders 23,13 respectively of the other channel. The predictor forms a weighted sum with 2' 1' e three lines.

weighting factors SO, S S for the three Although not shown in Fig. 1 each of these weighted signals will normally be a weighted sum of several samples delayed by 1, 2, 3... sample periods, i.e. in general they are not merely single weighting factors. As illustrated in Fig. 4, weighting factor Sn for the nth previous line will comprise individual weighting factors. sn1 sun.2 etc. by which successive samples from delay elements C (one clock period) are multiplied prior to being continued in an adder A.

The decoding process is substantially similar to the coding process, the two quantised difference signals being extracted from the transmitted signal and being added to predicted values to regenerate the odd and even line signals which can then be compressed in time and interleaved with an arrangement similar to the line store arrangement shown in Fig. 1. That part of the coder of Fig. 1 shown within the dotted lines may, in fact, be used as a decoder by feeding the quantised difference signals to points X and obtaining the regenerated signals from points Y.

The arrangement depicted in Fig. 1 differs from conventional video DPCM coders in that the video signal is divided into two channels, as described.

The predictor coding arrangements (units 11 to 1 6 and 21 to 26) operate in conventional manner, except that each of these arrangements obtains information concerning the preceding line from the other channel (via the half-line delay 16). It should also be noted that reference to one-line and half-line delays refer to a delay corresponding to one line (or half a line) at the point in question, although of course the actual time of the delay is equal to two, or as the case may be, one line period refered the input signal, due to the slowing down of the data in the line store arrangements 10,20.

It will be appreciated that, although the apparatus of Fig. 1 is more complex than a conventional coder, involving duplication of the predictor circuitry, a halving in the speed of operation of the predictor circuitry is obtained, which provides the possibility of using slower devices with resultant savings on cost and power consumption, or an increase in resolution.

The effect of this coding method on the resultant coded DPCM signal will now be considered. In Fig. 3 is shown an expanded view of the sample times: it is assumed that there are only ten samples in a line: although in practice there will be vastly more than this, this simplication does not affect the principle to be illustrated. In this figure, the tens digit denotes which line the sample is in and the one digit denotes its place in the line. As the prediction for any given sample is made at the time that it is required, it is apparent that the sample can be predicted only in terms of samples which have already occurred in time. Samples to be used in a prediction can be taken from the "present" line and from previous lines.Previous line samples are always available in time unless it is required to use samples half a line "to the right" (e.g. using sample 1 5 in the prediction for sample 20): this is not in any sense a limitation on the technique although it does impose a limitation on the ultimate degree to which the incoming signal can be "divided down". In so far as samples from the "present" line are concerned, two cases must be considered, namely where the predicted and prediction samples do or do not lie on opposite sides of a line boundary (in this respect "same line" prediction more accurately means the samples used in the prediction are much less than a line length from the predicted sample).Where a line boundary is not crossed, once again the samples are available-for example to predict sample 25 one might use samples 24,23 which are available. However, to predict sample 20 it is usual to use sample 1 9 (and possibly 18,17 ..... ), but in the present system this is not possible since, although these samples have obviously occurred, they have not yet been read out from the second store. This limitation is overcome by using instead the last few samples of the last line but one (i.e. they are a line late).

For the case under consideration this would be samples 9,8 . . since the odd lines are processed in one DPCM arrangement and even lines in another. Thus, the output signal obtained from the apparatus of the present invention differs from that obtained in the conventional arrangement only at the line boundaries. Whilst it may be argued that sample 9 is not a "correct" prediction for sample 20, it can equally well be argued that neither is sample 1 9 as this will be derived from the other side of the picture (or from the line blanking period). Thus, although the output signal will be different from that obtained in a conventional system, the degradation of the signal at line boundaries will not be significantly worse than that already suffered by conventional DPCM arrangements.

Although in many situations (e.g. transversal filter implementations) it is possible to increase the throughput rate by duplicating the circuitry and having dual circuits operating in parallel, combining (by interleaving) the processed samples in the appropriate order at the output end, this is not normally possible in recursive systems such as DPCM. However, the present invention recognizes that the nature of the video signal is such that the signal may be subjected to demultiplexing providing that the "boundaries" lie between lines of the video signal.

Although the systems shown involves treating alternative lines of the video signal in separate channels, it will be appreciated that it is equally possible, using the same principle, to divide the signal into three or more channels for parallel processing.

Claims (6)

Claims

1. A method of coding a video signal, comprising distributing the individual lines of the video signal cyclically between two or more processing channels, and retiming and encoding the signals in each of the channels by the use of a recursive coding procedure.

2. A method according to claim 1, in which the individual lines of the video signal are distributed alternately between two processing channels.

3. A method according to claim 1 or 2, in which the encoding in each channel is carried out being differential pulse code modulation.

4. A method according to claim 1, 2 or 3, in which the encoding step in each channel takes into account information from another (or the other) channel relating to the preceding line.

5. A method according to claim 1 substantially as herein described.

6. An apparatus for coding a video signal comprising means arranged in operation to distribute the individual lines of the video signal between two or more processing channels and to retime the signals to reduce the signal rate thereof for encoding, recursive encoding means connected in each channel, and means for combining the outputs of the encoding means.

6. An apparatus for coding a video signal comprising means arranged in operation to distribute the individual lines of the video signal between two or more processing channels and recursive encoding means connected in each channel.

7. An apparatus according to claim 6 having two processing channels.

8. An apparatus according to claim 6 or 7, in which each of the said recursive encoding means is a differential pulse code modulator.

9. An apparatus according to claim 8, including inter-connections between the said encoding means whereby each encoding means is supplied with signal information occurring, in the input video signal, one line period earlier.

10. An apparatus according to any one of claims 6 to 9, in which the distribution means comprises a respective store for each channel and control means arranged for entering successive lines into respective stores and reading them out at a lower rate.

11. A video signal coding apparatus substantially as herein described with reference to the accompanying drawings.

1 2. A decoder for use with the encoder of any one of claims 6 to 11, comprising inputs for receiving encoded signals of respective chann'els, recursive decoding means for each channel, and means for combining the decoded signals.

New claims or amendments to claims filed on 5th November 1 982 Superseded claims New or amended claims:

1. A method of coding a video signal, comprising: distributing the individual lines of the video signal cyclically between two or more processing channels; in each channel, retiming the signals to reduce the signal rate thereof for encoding and encoding the signals using a recursive coding procedure; and combining the encoded signals from the individual channels.

3. A method according to claim 1 or 2, in which the encoding in each channel is carried out using differential pulse code modulation.