GB2166021A - Transmitting television signals - Google Patents

Abstract

Transmitting television signals in sequential format within a restricted transmission bandwidth from an input video signal 12 consisting of an interlaced video signal is achieved by storing 10 the interlaced signal and sequentially sending a proportion (e.g. half) of the lines in a frame to a two-dimensional filter 14, such as a transversal filters. The filter discards part of the spatio-temporal spectrum of the signal which contains relatively little picture information. The output of the filter is sub-sampled 15 and lowpass filtered 16, and timing information is added, 19, before transmission. At the receiver (Fig. 4), the signal is applied to a re- sampler and a two-dimensional filter and thence to a picture store which holds lines from which the output signal is interpolated. A signal generated at 1125 lines per picture and 80 fields per second with a 2:1 interlace can be transmitted at 562 lines per picture and 40 fields per second non-interlaced. <IMAGE>

Description

SPECIFICATION Transmitting television signals This invention relates to the encoding and decoding of television signals and more particularly to the encoding and decoding of a high definition television signal having a restricted transmitted bandwidth.

Currently much effort is being devoted to the identification of the optimum parameters for a worldwide high definition television (HDTV) studio standard. In particular, the EBU have commended one possible set of parameter options for detailed study, see EBU document GT V 189, 'EBU comments on the HDTV production standard' contributed to CCIR Interim Working Party 11/6 (1984).

However, in practice it is difficult to divorce entirely the requirements of transmission and display from those of signal origination, since it is important to ensure that realistic prospects for distributing a given HDTV standard do, in fact, exist. It is also likely that the choice of transmission and display standards that are related by simple factors to the production standard (e.g. the source of camera standard) will lead to the minimum level of perceived artefacts.

It has already been proposed to use sophisticated 'movement adaptive processing' in television receivers, but experience of these has led to the conclusion that the results produced by such systems do not justify their complexity.

The total bandwidth occupied by a suitable HDTV production standard is likely to be substantial. The 1125/80/2:1 standard suggested in the above EBU document would, for example, use a bandwidth of over 30 MHz for the luminance signal alone, even if line and field blanking were removed. The designation 1125/80/2:1 means that there are 1125 lines per picture, 80 fields per second, and two interlaced fields per picture, i.e. a 2:1 interlace. Basic information theory demands that, if this high figure is reduced to a significantly lower level to allow transmission over a limited bandwidth channel, there must be occasions when distortions of the original HDTV picture occur. In general, the more sophisticated the bandwidth reduction technique the less frequent are the distortions but experience indicates that they are more objectionable when they do occur.Two approaches are therefore possible. Either a very sophisticated technique of redundancy reduction may be used, to try to minimise the occurrence of picture distortion, or else a simple and predictable method may be empioyed and the parameters adjusted to give an acceptable level of final picture quality; this second approach will not give as high a picture quality, under ideal conditions, as the first but neither will it suffer the catastrophic failures to which the more sophisticated method may be prone.

Previous work on display up conversion has indicated that it is extremely difficult to generate a higher field rate, or sequentially scanned, display signal from an interlaced transmission signal, (see BBC Research Department Report No. 1983/8, The improved display of 625-line television pictures, ROBERTS, A.). Sophisticated adaptive interpolation is required and, while this can be made to operate satisfactorily on test signals, typical television scenes cannot yet all be satisfactorily processed. In contrast, however, the sequentially scanned signals originating from film are relatively easily handled every day in most broadcasting organisations, the main defect being the objectionable level of judder visible on moving objects.

We have also found that interlaced television cameras cannot take full advantage of the total vertical resolution implied by the number of scan lines in a complete television picture.

The vertical spatial frequency response of a typical present-day 625-line television camera falls off rapidly above about 156 cycles per picture height. Future HDTV cameras are unlikely to be significantly different, especially since vertical resolution can only be improved at the expense of an increased temporal integration.

We have realised that, while current proposals for HDTV systems provide for increased resolution, the increase obtained is often substantially less than the attendant increase in signal bandwidth, the balance of the increased bandwidth being used to reduce the effect of scanning impairments. We have appreciated that the use of interlace in a source standard, although desirable for a fullbandwidth transmission system, works against the use of more sophisticated methods of scanning products to improve the quality of displayed pictures in a practical HDTV transmission system.

The present invention provides apparatus for encoding an interlaced video signal into a sequential video signal of restricted frequency bandwidth, the apparatus comprising input means for receiving an interlaced video signal; selection means connected to the input means for providing sequentially a selected proportion of the lines in each frame of the input interlaced video signal; twodimensional filter means for receiving the said selected lines and for discarding one part of the spatio-temporal spectrum of the said lines which part contains less picture information than another part of the said spatio-temporal spectrum; a sub-sampler connected to an output of the two-dimensional filter means for sub-sampling the said output; and synchronising means for adding to the output of the subsampler timing information which is related to the sampling of the sub-sampler.

Thus an input 1125/80/2:1 signal may be processed to provide a signal at a 562/40/1:1 standard for transmission (the designation 1:1 indicating a non-interlaced, or sequential, signal).

At the receiver the 562/40/1:1 signal is converted back to 1125/80/2:1 for display; an alternative display standard, but technically harder to achieve, could be 1125/80/1:1. The conversion process requires that a complete transmitted field be stored in the receiver, but the interpolation is relatively simple.

In the encoding, those portions of the spatio-temporal spectrum of the production signal that are less completely filled by the studio camera are discarded, while the transmitted signal is of sequentially scanned type so as to allow an easy up-conversion at the display.

One simple way in which this can be achieved is to transmit only one of the two interlaced fields. This introduces alias components into the signal which would be embarrassing in a production standard; these aliases can actually be beneficial in a transmission standard, however, as they can improve the subjective resolution of the final display.

Thus, the present invention further provides a method of encoding an interlaced video signal into a sequential video signal, the method comprising two-dimensionally filtering a proportion of the lines in a frame of picture signal to discard one part of the spatio-temporal spectrum of the said signal which contains less information than another part of the spatio-temporal spectrum; subsampling the result of the filtering process and adding timing information to the output of the sampling process which is related to the sampling process.

The invention also provides methods of encoding, transmitting and decoding video signals as defined in the appended claims.

The invention will now be described in more detail by way of example, with reference to the drawings, in which: Figure 1 is a block diagram of an encoder according to the invention; Figure 2 is a diagram illustrating the twodimensional passband of a diagonally-filtered 1125/80/2:1 signal, with the passband of the 562/40/1:1 transmission shown for comparison in dashed lines; Figure 3 shows diagrams at (a) illustrating a quincunx subsampling structure for the 562/40/1::1 transmission standard, at (b) the two-dimensional (vertical/horizontal) passband of the quincunx structure of Figure 3(a), at (c) the equivalent two-dimensional passband of an interlaced television signal, and at (d) an alternative frame-interlace quincunx sub-sampling structure,; Figure 4 is a block diagram of a decoder according to the invention; Figure 5 is a diagram illustrating the interpolation which can be used to convert from 562/40/1:1 to 1125/80/2:1 standard at the receiver.

Referring to Figure 1, a complete field of conventional picture information is stored in a picture store 10 and each element of picture information assigned an address from an address generator 11. The output of the picture store 10 is programmed to send every other line of picture information. Thus, an incoming interlaced 1125/80/2:1 television signal 12 is sent out of the picture store 10 as a 562/40/1:1 sequential signal 13 at halfspeed. Clearly, the address generator needs to be synchronised to the 1125/80/2:1 signal 12. Details of the synchronising circuitry have been omitted for the sake of clarity.

The output of the picture store 10, the 562/40/1:1 television signal 13, is then fed to a two-dimensional filter 14, which can be a standard transversal filter using a combination of line and element delays, which removes diagonal spatial frequencies as shown in the graph in Figure 2. Figure 2 shows the passband of a diagonally-filtered 1125/80/2:1 signal bounded by the full lines, forming the shape of a diamond, and the passband of the corresponding 562/40/1:1 signal in dashed lines. The axes are calibrated in cycles per picture height (c/ph) and cycles per picture width (c/pw) as shown.

The bandwidth required by the 562/40/1:1 signal 13 is still too high to allow it to be transmitted via a single 27 MHz WARC channel, as described in the Final Act of the World Administrative Radio Conference for the Planning of the Broadcast-Satellite Service, Geneva 1977 (International Telecommunications Union). However, advantage can now be taken of its sequential structure to allow a quincunx sub-sampling technique to be applied, as in Figure 3a, at a sub-sampler 15 in Figure 1.

This further reduces the bandwidth by a factor of up to 2:1, using a sub-Nyquist sampling frequency, i.e., one less than twice the highest significant frequency in the video signal.

But this reduction is made at the expense of removing some of the diagonal spatial components of the wanted signal, which contain relatively little picture information. However, the sequentially-scanned nature of the transmitted signal allows a better compromise to be achieved than is possible with the present day interlaced system.

The two-dimensional spatial frequency 'passband' achievable with the sampling structure of Figure 3a is shown in Figure 3b. If the transmission structure had been interlaced, each individual field would have had to have been sampled in a quincunx structure, giving the spatial frequency 'passband' shown in Figure 3c; this has a much higher subjective loss of resolution than that of Figure 3b. As a further alternative the quincunx sampling structure could be arranged to interlace on successive pictures, as shown in Figure 3d. Here, the circles represent the elements transmitted on the first transmitted frame and the crosses represent the elements transmitted on the second transmitted frame.While this would improve the static resolution, the benefits may not be worth the penalty of re-introducing an interlace component into the transmitted signal, since the problems of choosing a good method for display up-conversion will almost certainly recur. This quincunx sub-sampling is in effect a weighting process in which each sampled element is weighted according to the value of the elements surrounding that sampled element in a quincunx arrangement, thus effectively further two-dimensionally filtering the video signal.

In order to calculate the bandwidth requirements for transmission it is necessary to know the number of active lines present in the incoming 1125/80/2:1 camera signal 12, assuming that line and field blanking are not transmitted. Taking a value of 1060 lines, the transmitted line rate becomes: 1060 X40=21,200 Hz.

2 Tests have shown that there is a slight loss of vertical resolution in the transmission signal. If the horizontal resolution is also reduced slightly to match, a further saving in transmission bandwidth can result, assuming a horizontal resolution limit of 525 cycles per active line, as indicated in Figure 2. This figure assumes a reduction of the HDTV production signal horizontal resolution by about 25%--estimated as of the same order as the noted reduction of subjective vertical resolution. The luminance bandwidth becomes: 21,200 X 525 = 11.13 MHz.

The output of the sub-sampler 1 5 is then lowpass filtered by a filter 16 the output of which forms the final video signal 17.

In order that the phase of the decoder sampling matches that of the coder sampling, synchronising information in the form of a short burst on every video line, produced by a sampling pulse generator 1 8 for example, must be added to the signal before transmission at an adder 19. The circuits may be either analogue (using a CCD device to store the incoming picture for example) or digital. In the latter case an analogue to digital converter must be added at the input and a digital to analogue converter at the output or at any other convenient point in the signal chain such as between the subsampler 15 and the lowpass filter 16 for example.

After quincunx sub-sampling and lowpass filtering the luminance bandwidth falls to just under 5.6 MHz. Assuming alternate-line transmission of the chrominance signals, coupled with a horizontal chrominance bandwidth half that of the luminance signal, the total vi deo signal bandwidth becomes 8.4 MHz.

When audio and synchronising signals are added the overall bandwidth rises to 9.0 MHz; this figure assumes that only two sound chan nels are required, that the same data rate is used as for C-MAC transmission, and that the same total time is used for transition and clamp periods as in the C-MAC system, i.e.

because of the slightly higher line rate of this HDTV transmission system, slightly less time is allowed per line.

Operations in the decoder, illustrated in Figure 4, are very similar to those in the encoder. After regeneration of the correct phase of the sampling pulse 20, which is achieved by deriving the timing information in a separator 21 from the incoming sequential signal 22 and used to drive a sampling pulse generator 23, the signal is re-sampled by a re-sampler 24, which is identical to the sub-sampler 15 in Figure 1, to form the correct alias structure, and then re-filtered by a matching two-dimensional filter 25. The resultant diagonally-filtered signal 26 is then fed to a picture store 27 (again with a minimum capacity of one field of the original 1125/80/2:1 signal). The store has four outputs 28 operating at twice the speed of the incoming 562/40/1:1 signal.

These four outputs feed an interpolator 29 which generates the final 1125/80/2:1 signal 30. Purely vertical interpolation is used, as illustrated in Figure 5. This potentially gives rise to errors in the converted signal, but in practice these are not significant. The circles represent incoming lines at 562/40/1:1 and the crosses represent the outgoing lines at 1125/80/2:1. An address generator 31 for the picture store 27 can also generate control information for the interpolator 29 and, as in the case of the encoder, will require some information to be sent with the video signal for synchronisation purposes. Once again this is not shown for the sake of clarity, but could be very similar to the synchronisation pulses routinely used in 625-line television. As in the coder, the decoder may use either analogue or digital circuitry or a combination of both.

The basic system may be enhanced, for example as follows.

The static resolution of an HDTV picture is impaired in three ways by the transmission system described above. The 80 Hz interlaced/40 Hz sequential/80 Hz interlaced conversion reduces the vertical resolution and introduces some additional 'knotting' on diagonals, and the 2:1 quincunx down-sampling further reduces diagonal resolution.

It seems likely, however, that the capacity of a WARC BS1977 channel will be higher than the 9 MHz required by the basic transmission system. Use of this additional bandwidth would allow the static resolution of the transmitted picture to be increased by sending a low field rate detail signal, in a similar manner to that proposed by Glenn et al, 'Reduced bandwidth requirements for compatible high definition television transmission', EBU document GT V1/HDTV 123, (1984).

The best format for such a signal will need to be investigated but, in order to gain insight into the possibilities, one suggestion would be to form a diagonally-filtered version of the 1125/80/2:1 input signal by combining successive odd and even fields (the resulting temporal impairment being insignificant since only the low temporal frequencies will be transmitted). Such a signal will occupy a luminance bandwidth of approximately 15 MHz (depending on the exact details of the lowpass filtering used); the excess bandwidth on stationary pictures, when compared with the basic HDTV transmission system, is then 9.4 MHz. Reduction of the temporal frequency of this additional detail by a factor of 4:1 (i.e. to a sequentially-scanned field rate of 10 Hz) will then compress the bandwidth to 2.35 MHz.A greater degree of compression could be used for the chrominance information resulting in a total bandwidth for the slow scan detail signal of around 3 MHz. The total bandwidth required from the WARC channel would then be around 12 MHz.

One potential problem with using a slow scan detail signal is that it might leave behind high frequency edge information when objects move. This can be avoided by using an adaptive transmitter to switch off the detail signal in areas of movement; it is not necessary for the receiver to be adaptive. Such an approach not only keeps the receiver simple but also provides the maximum possibility of adequate detection of motion, since this can be accomplished on the 1125/80/2:1 signal 12 before bandwidth compression.

A second possible enhancement is concerned with loss of resolution in moving areas, due to combing of the basic system, which is most noticeable on gross movements of the whole scene, such as during a pan; movements of small parts of the scene are much less degraded. Measurement of the direction and speed of any such bodily motion, either by processing the electrical signal from the camera or by fixing a suitable accelerometer to the camera head (see U.K. Patent Application 2,116,397A), and transmitting this information to the receiver would allow an improved interpolation to be used for the display up-conversion. Ideally, the mean motion vector of any given scene would be transmitted, so that combing is only present on any differential motion components.

It is worth noting that there is no theoretical conflict in the use of such a vector compensation scheme because of the sequential nature of the transmitted signal. Each transmitted field is complete. In contrast, the vector compensation of a sub-sampled signal can present significant probiems if the distance moved between successive fields is not close to a multiple of the subsampling pitch.

In summary, the system described is based upon the use of an 80 fields per second, interlaced studio standard and operates by transmitting one field of an interlaced pair (i.e.

a transmission standard of 40 fields per second, sequential). The basic system is nonadaptive, which has the advantage of allowing relatively simple and reliably predictable receivers to be used; some suggestions for possible compatible enhancements have also been made, however, which might allow a range of possible reception options.

While the system has been described in relation to a 1125/80 standard, it can be used with other potential standards such as 1249/80 for example. It is however expected to be of most practicality with source field rates that are substantially higher than the conventional 50 or 60 Hz, such as 75Hz or greater, and a source line rate higher than the conventional 525 or 625 lines per picture, such as 800 lines or greater.

Claims (14)

1. Apparatus for encoding an interlaced video signal into a sequential video signal of restricted frequency bandwidth, the apparatus comprising: input means for receiving an interlaced video signal; selection means connected to the input means for providing sequentially a selected proportion of the lines in each frame of the input interlaced video signal; two-dimensional filter means for receiving the said selected lines and for discarding one part of the spatiotemporal spectrum of the said lines which part contains less picture information than another part of the said spatio-temporal spectrum; a sub-sampler connected to an output of the two-dimensional filter means for subsampling the said output; and synchronising means for adding to the output of the sub-sampler timing information which is related to the sampling of the subsampler.

2. Apparatus as claimed in claim 1, wherein the sampled element is weighted according to the values of the elements surrounding the said sampled element in a quincunx arrangement.

3. Apparatus according to claim 1 or 2, wherein the selection means comprises a picture store for retaining at least a frame of the input interlaced video signal and transmitting a predetermined selection of the lines therein to the two-dimensional filter means.

4. Apparatus as claimed in claim 1, 2 or 3, wherein every other line of a frame of the input interlaced video signal is supplied to the two-dimensional filter means.

5. Apparatus as claimed in any of the preceding claims, including a low pass filter connected with the output of the sub-sampler.

6. Apparatus as claimed in any of the preceding claims, wherein the input interlaced video signal comprises 1125 lines per picture at 80 fields per second and the thus encoded sequential signal comprises 562 lines per picture at 40 fields per second.

7. Apparatus for decoding a sequential video signal of restricted frequency bandwith, the apparatus comprising: a separator for receiving the input sequential video signal and isolating the timing information therefrom; a resampler connected to the separator and timed by the isolated timing information, for sampling the video signal in order to substantially reconstitute an unsampled sequential video signal; two-dimensional filter means, for receiving the output of the resampler and substantially regenerating one part of the spatio-temporal spectrum of the said lines; a picture store, connected to the output of the filter means for transmitting the filtered resampled signal through a plurality of outputs; and interpolator means for receiving the plurality of outputs of the picture store and reconstituting an interlaced video signal therefrom.

8. Apparatus as claimed in claim 7, wherein the contents of the outputs of the picture store are governed by a pre-programmed address generator which simultaneously times the interpolator.

9. Apparatus as claimed in claim 7 or 8, wherein the interpolator weights each element according to the corresponding elements in preceding and subsequent lines in the frame.

10. A method of transmitting a video signal comprising: producing a source video signal at a source line rate, a source field rate, and in 2:1 interlace form; deriving a transmission signal at a transmission line rate which is half the source line rate, a transmission field rate which is half the source field rate, and in sequential form; transmitting the transmission signal; and at a receiver up-converting the transmitted signal to at least the source line rate and source field rate.

11. A method according to claim 10, wherein the source line rate is 800 lines per picture or greater.

12. A method according to claim 10 or 11, wherein the source field rate is 75Hz or greater.

13. A method of encoding an interlaced video signal into a sequential video signal, the method comprising: two-dimensionally filtering a proportion of the lines in a frame of picture signal to discard one part of the spatio-temporal spectrum of the said signal which contains less information than another part of the spatio-temporal spectrum; sub-sampling the result of the filtering process; and adding timing information to the output of the sampling process which is related to the sampling process.

14. A method of decoding a sequential video signal encoded according to the method of claim 13, the method comprising: separating the sequential signal from timing information in the incoming signal; resampling the output of the separator and timing the resampling by means of the separated timing information in order to substantially reconstitute the unsampled sequential video signal; two-dimensionally filtering the resampled video signal in order to substantially regenerate the one part of the spatio-temporal spectrum of the said lines which was discarded whilst being previously encoded; and interpolating from the two-dimensionally filtered output to reconstitute the interlaced video signal therefrom.