GB2138238A - High definition video signal transmission - Google Patents

Abstract

A high definition video signal, such as a 1249 line 2:1 interlaced signal is transmitted through a main channel and an auxiliary channel, the main channel being compatible with a conventional (625 line) video signal. The high definition signal is separated in frequency space into the two transmitted signals namely to high and low pass bands relative to the repetition frequency of the signal. In a first example (Figures 1-2) the horizontal frequency of a digital input signal is the criterion, low frequency components being transmitted in the main channel and high frequency components transmitted in the auxiliary channel, after frequency conversion. These component signals are sampled at twice the 625 line signal sampling frequency and main and auxiliary signal samples taken alternately in time. In a second example the main channel takes low frequencies (11, 12, 13, 14) defined in the vertical frequency/temporal frequency plane and the auxiliary channel takes down-converted high frequencies (21, 22, 23, 24) in the said plane. <IMAGE>

Description

SPECIFICATION High definition video signal transmission This invention is concerned with systems for transmitting and receiving a video signal having increased definition compared with the conventional broadcast standards currently in use. Such a video signal may be termed an HDTV (high definition television) signal.

The invention can in principle be applied to component signals or any composite signal in which the components are line simultaneous but not on a carrier (or on a line-locked carrier).

The invention provides a method of transmitting an input video signal having a relatively high definition format over two channels namely a main channel signal which is compatible with a standard format video signal, and an auxiliary channel, in which the high definition input signal is filtered with a low passband in relation to repetition frequencies of the signal to provide the compatible main channel signal, and the high definition input signal is filtered with a high passband in relation to repetition frequencies of the signal to provide the auxiliary channel signal.

Additional features of the invention are defined in the appended claims.

A particularly preferred system embodying the invention is concerned with an HDTV source scanning format (e.g. from the camera) which has a 2:1 interlace, with nominally double the number of lines compared with a conventional format. For example, if the conventional format were 625/502:1, the source format would be 1249/50 2:1. There is a requirement that such a signal should be transmitted through two video-bandwidth channels. The first, main, signal would be a 625/50 2:1 compatible signal; that is to say this signal alone could be received by a conventional receiver and displayed at normal definition. On the other hand an HDTV receiver would receive in addition the second, auxiliary, signal and use this to increase the definition of the displayed signal.

Embodiments of the invention will be described by way of example with reference to the drawings, in which: Figure 1 illustrates on the left-hand side of the Figure various frequency spectra involved in a first embodiment of the invention which discriminates on the basis of horizontal frequency, and on the right-hand side of the Figure corresponding digital sample structures; Figure 2 is a block circuit diagram of the first embodiment of the invention; Figure 3 shows the baseband of a conventional video signal in the vertical frequency/temporal frequency plane; Figure 4 is a similar diagram for the input high definition signal; Figure 5 defines portions of the baseband area in Figure 4; Figure 6 is a plot of vertical position against time, showing the positions of lines in the high definition input signal and in the output signals;; Figure 7 shows the areas in the frequency plane selected for the main channel in a second embodiment; Figure 8 shows the spectrum of the main channel signal of the second embodiment after subsampling; Figure 9 illustrates the basic passband for the auxiliary channel signal; Figure 10 shows the high vertical/temporal frequencies selected for the auxiliary channel signal in the second embodiment; Figure 11 shows the auxiliary channel frequency spectrum in the vertical/temporal plane after sub-sampling; Figure 12 shows the main channel signal at the receiver with zero-value samples added; Figure 13 shows the auxiliary channel signal at the receiver with zero-value samples added; Figure 14 shows the spectrum of the combined up-converted main and auxiliary channel signals at 4N+1 sequential format;; Figure 15 shows the spectrum of the combined signal of Figure 14 after selecting alternate samples to give a field-interlaced output; Figure 16 is a schematic block circuit diagram of the transmitter of the second embodiment; and Figure 17 is a schematic block circuit diagram of the receiver of the second embodiment.

The systems to be described assume the availability of a source which provides a video signal having 4N+1 lines per picture with a 2:1 interlace, as compared with a 2N+1 line conventional signal. Thus the high definition source signal may be a 1249/50 2:1 signal for example. The signal for this source is to be prepared for transmission in two channels each of the bandwidth of a conventional video signal. The first channel is the main signal and is compatible with a conventional 2N+1 2:1 signal (e.g. a 625 line signal), to the extent that it can be received and displayed by a conventional receiver. The second, auxiliary channel carries the extra detail, that enables a high definition receiver to produce a picture of higher quality.

The first embodiment to be described separates the signals in the (horizontal) frequency spectrum, whereas the second embodiment separates the signals in the vertical frequency/time domain, as will become apparent. The high frequencies in each case which are not transmitted in the main channel are down-converted for transmission in the auxiliary channel.

Reference will first be made to Figures 1 and 2 to illustrate the first embodiment of the invention. Figure 1 shows various frequency spectra on the left-hand side of the Figure, and corresponding sample structures on the right-hand side. Figure 2 is a block circuit diagram of a transmitter and receiver system embodying the invention.

The spectrum of a signal sampled at a frequency f,, together with its sample structure, is shown at (a) in Figure 1. The sampling operation causes the baseband spectrum to repeat at multiples of f,. The present requirement is to transmit this signal along two channels, the main and auxiliary channels, each of which can carry samples at a rate of fs, the main signal furthermore carrying a signal which is compatible with a conventional signal.

The high definition signal received from the signal source at input 10 is first pre-filtered with a digital filter 12 having a low-pass form and cutting at frequency fas. This characteristic is shown at (b) in Figure 1 and will repeat at multiples of f3. The +'s on the right-hand side of Figure 1 show the positions in time of the thus-modified samples. Then alternate samples are selected in a circuit 14 to produce the spectrum shown at (c) in Figure 1, in which interleaved spectra centred on odd multiples of fs are produced. These samples, representing the 0 to 1/4f, band of the spectrum shown at (c), form the signal in the main or compatible transmission channel 16.

As can be seen the function of the prefilter 12 is to prevent significant overlap between the two spectrum parts, although this has been exaggerated at (c) for clarity.

At the same time, the original input signal is applied to a digital high-pass filter 22 to pre4ilter the signal with a high-pass characteristic cutting at fas, as shown at (d). Again the + signs indicate the modified samples. From these samples the interlaced alternate samples are selected by a circuit 24 to produce the spectrum shown at (e). This interlaced half-rate sub-sampling causes the interleaved spectra to be inverted, and these are shown in dashed outline at (e). These samples, representing the fs to fs band at (e), constitute the down-converted high-pass filtered signal transmitted in the auxiliary channel 26. Again, the function of the prefilter 22 is to prevent overlap of the spectra.

The interlaced or interleaved sampling shown at (c) and (e) in Figure 1 is such that the samples of the main signal are taken at (say) the odd-numbered sampling points at the fs sampling rate, whereas the samples forming the auxiliary signal are taken at the even-numbered sampling points.

At the receiving end, zero-value sample insertion circuits 32 and 42 insert alternate zero-value samples into the incoming main and auxiliary sample streams, as shown at (f) and (g) in Figure 1. The zero value samples are illustrated by 0's on the right-hand side of the Figure. The addition of these zero-value samples does not alter the spectra, but does allow the signals to be filtered with digital filters based on the original sampling frequency f,. The main-channel signal is thus post-filtered with a low-pass form of filter, cutting at fas, to produce the spectrum shown at (h) in Figure 1.This filtering in circuit 32 largely removes the spectral energy centred on odd multiples of fas. The post-filtering of the high-frequency signal in circuit 42 has a high-pass characteristic cutting at 1/4f,, to produce the spectrum shown at (i) in Figure 1. This filtering largely removes the spectral energy centred on multiples of f,. Each sample insertion circuit and associated post-filter may be regarded as an interpolation circuit.

Finally, an adder 40 adds the two up-converted sample streams together to yield on line 50 an output signal having substantially the original spectrum, with some impairments. These impairments are of two types. Firstly, the wanted spectrum is modified in the transition region between the low and high pass regions. Secondly, vestiges of the spectra centred on odd multiples of 1/2fs remain. However, it can be shown that, with a suitable choice of pre- and post-filters, these impairments can in principle be eliminated. Figure 1 shows examples of such filters for it can be seen that the spectra of (h) and (i) will combine to reconstruct the spectra at (a) perfectly, whilst cancelling the interleaved vestiges. This is made possible only by the interlaced sub-sampling of the high-pass signal.

Specifically, if the low and high pass characteristics of pre-filters 12,22 are F11 and F12 respectively, and the low and high pass characteristics of postfilters 34,44 are F21 and F22 respectively, then the conditions for avoiding impairment are: F17F21 + F12F22 = 1 and F21 F11 - F22F12 = 0 where F11 etc. denotes the characteristic F11 etc. shifted by 1/2fas. It can be shown that this can be satisfied if (i) the post-filter is identical to the corresponding prefilter: i.e. F22 = F72 and F21 = F11 (ii) the characteristic of the high-pass filter is the reflection of the characteristic of the low-pass filter about f3, i.e. F12(1/2f8 - v) = F1(v) and and (iii) F11 is 'square-root anti-symmetrical' about 1/4fas i.e. F112(v) + F112(1/2fs - y) = 1.

Within these constraints there is still an infinite variety of filter characteristics that can be used.

It will thus be seen that the system of Figures 1 and 2 includes a high-pass filter to extract a down-converted high-pass signal and a high-pass interpolator to provide an up-converted high-pass signal at the receiver. It should be noted in particular that the interlaced sub-sampling of the high pass signal is important in that this is required to enable the vestigial spectral impairments of the reconstituted signal to be eliminated by a suitable choice of filters.

At the receiver both the received signals are up-converted to the f, sampling rate prior to digital filtering.

Similar principles to those employed in the one-dimensional horizontal frequency case can be applied to a two-dimensional embodiment in which the down- and up-conversion takes place between interlaced sampling structures having the same field period but differing line pitches by a factor of two.

Again it is assumed that the source scanning format is 2:1 interlaced with nominally double the number of lines as before. The first or main channel carries a conventional 2:1 interlaced signal and the second or auxiliary channel carries the increased vertical/temporal detail on an identical scanning format.

In this example the main channel carries the low vertical and temporal frequencies that would be found in a conventional signal, whereas the auxiliary channel carries the high vertical and temporal frequencies associated with the source format that lie outside the range carried by the main signal. Again there is a down-conversion of format at the transmitting end to produce the two transmitted signals and an up-conversion at the receiving end to change the format of the signals to the source format before they are combined to reproduce the original source signal.

Figure 3 shows the vertical/temporal spectrum of a conventional signal, assuming an idealised baseband spectrum, sampled at 2N+1 lines per picture. Thus we are here concerned with line-wise sampling, or sampling in the vertical direction and field-by-field or 'temporal' sampling, rather than sampling in the horizontal direction. (The type of diagram shown in Figure 3 is used and explained in BBC Engineering, Vol 104, September 1976, pages 8 to 39.

As can be seen, the scanning action repeats the spectrum in two dimensions on a regular quincunxial lattice, the repeat units along the axes being the number of lines per picture and the number of fields per second. Figure 4 shows the vertical/temporal spectrum of the higher definition signal, from the signal source. This signal is sampled at the source standard of 4N+1 lines per picture 2:1 interlaced, with the same field rate F as the conventiona signal. As can be seen the area occupied by the baseband spectrum is twice as large as in Figure 3, indicating that the source standard is capable of supporting twice the vertical bandwidth.

In order to transmit the full resolution of the source signal, the source baseband signal can be partitioned into low and high frequency areas. Figure 5 shows how the baseband spectrum of the higher definition format can be subdivided into the low-frequency spectrum of the conventional format (areas 11, 12, 13 and 14) and the spectrum of the extra details (areas 21,22,23 and 24). These two spectra have identical areas and so can be sent down identical channels.

The source signal with the spectrum of Figure 4 is first subsampled on the conventional format using an interpolation aperture spectrum corresponding to the low-pass area. This produces a signal with the spectrum of Figure 3 where the baseband spectrum corresponds to the areas 11,12, 13 and 14. In orderto transmit the low-frequency area, the source signal is thus first notionally, up-converted to a 4N + 1-line sequential standard by inserting alternate zero-valued samples as shown in Figure 6. This up-conversion is necessary because the conventional 2N+1/2:1 scanning format has lines at positions on the vertical distance/time diagram of Figure 6 (shown by +'s) where there are no lines in the 4N+1/2:1 format (these being shown by X's).Thus it is necessary to up-convert to a notional standard which includes all the sample positions of both formats, namely 4N+ 1/1:1 format (or 4N+1 sequential). The samples are than pre-filtered with a low-pass form of digital filter which selects the low frequency area as shown in Figure 7, thereby creating samples at all the sites in Figure 6. Then sub-sampling at the channel standard takes place by selection of the + samples as shown in Figure 6, to create the spectrum shown in Figure 8. The up-conversion eliminates the interleaved spectrum centred on (2N+1, 1,1/2F) which would interfere with the baseband after subsampling. It is 'notional' in that it does not, in fact, take place, but is useful for the purposes of explanation.The process of up-conversion plus fixed filtering is eqivalent to filtering at the source rate with a varying algorithm. As can be seen from Figure 6, there are two sets of main channel samples, one set being coincident with the source samples and the other with zero-valued samples. Hence two different algorithms apply to the two sets, being complementary parts of the fixed filter. These channel samples form the main channel compatible signal.

In addition, the source signal is interpolated using the interpolation aperture spectrum derived from the baseband spectrum of Figure 9 and then subsampled on the conventional format. Although the aperture passes vertical frequencies only up to N+c/ph, the detail beyond this is carried in 'fold over' spectra, as noted below. Thus the source signal is up-converted to a 4N+1 sequential signal, as before, by the addition of zero-value samples (lines), and is pre-filtered with a high-pass form of digital filter which selects the high vertical/temporal frequencies as shown in Figure 10. This Figure is seen to be built up from the baseband spectrum of Figure 9.It will be noted in Figures 9 and 10that although there is a cut-off at nominally N cycles/picture height, the spectral components beyond N cycles1picture height are retained in the fold-over area belonging to adjacent repeated spectra in the source. Thus, for example, area 21 is divided in half, one half being in the 'normal' position and the other being in the fold-over position. This filtering operation produces samples at all the sites in Figure 6 as before. But now sub-sampling at the channel standard takes place at the interleaved sites, as shown by squares in Figure 6, to create the spectrum shown in Figure 11.

This is formed by repeating the spectrum of Figure 10 according to the parameters of the channel standard, but the interleaving of the sample sites inverts the spectra centred on odd multiples of N cycles/picture height. These are shown shaded, and the spectrum of Figure 11 is only partially built up for clarity. It can be seen that the central stop area created by the pre-filter (see Figure 9) is filled by the spectrum of the adjoining layer.

At the receiver main signal is upconverted to the higher standard using the same low-pass interpolation aperture spectrum as at the transmitter. Similarly the detail signal is upconverted using the same high-pass aperture spectrum as at the transmitter. The spectral areas are now re-ordered so as to restore the detail spectra to their correct places, and finally the two signals are added together to yield a signal with full definition.

More particularly, at the receiving end the incoming samples are interleaved with zero-valued samples, notionally on a 4N+1 line sequential site structure, as shown in Figures 12 and 13. The main signal, still having the spectrum of Figure 8, is then post-filtered with a low-pass filter, similar to the pre-filter. This filtering largely removes the spectra centred on odd multiples of nominally N and 2N cycles/picture height to yield a spectrum similar to that in Figure 7, except that there will be vestiges of the rejected spectra at the wanted spectra boundaries.

Similarly the high-pass signal, having the spectrum of Figure 11, is post-filtered with a high-pass filter, similar to the pre-filter. This filtering largely removes the spectral components centred on even multiples of 2N cycles/picture height and selects the high frequency components asymmetrically about odd multiples of N cycles/picture height to give a spectrum similar to that in Figure 10, except that there are, again, vestiges of the rejected spectra at the boundaries.

Adding the two up-converted sampled signals together produces a signal with the spectrum of Figure 14, together with impairments due to imperfect spectral joining and spectral vestiges, as in the one-dimensional case. Finally, selecting alternate samples on a field-interlaced basis as shown in Figures 12 and 13 yields the spectrum of Figure 15 which is identical to the original spectrum of Figure 4. The up-conversion to the full 4N+1 line sequential structure is, once again, notional in that it does not physically take place, the up-conversion being only to the 4N+1 line interlaced structure. As Figures 12 and 13 show, there are four possible phase relationships between the 4N+1 line interlaced samples and the incoming samples of each channel.Thus up-conversion involves four algorithms for each channel signal, each algorithm being a subset of the fixed algorithm corresponding to the filter having the characteristic of Figure 7 or 10.

Just as in the one-dimensional case, a suitable choice of pre- and post-filters, together with interlaced sampling of the high-pass signal which inverts spectra centred on odd multiples of N cycles/picture height, enables the impairments to be eliminated.

Block diagrams of apparatus to perform such operations are shown in Figures 16 and 17. At the sending end of Figure 16, the 4N + 1 line interlaced source signal at input 60 is down-converted through a low-pass interpolator 62 to a 2N+1 line interlaced compatible signal, the main signal. The interpolator algorithm is based on the principles described above. At the same time the source signal is down-converted through a high-pass interpolator 64 to a 2N+1 line interlaced high-pass or auxiliary signal, the sampling sites being interlaced with those of the main signal.

At the receiver, the compatible end auxiliary signals are up-converted via low-pass and high-pass interpolators 66 and 68 respectively to yield 4N+1 line interlaced signals which are then added in an adder 70 to form a signal ideally identical to the original signal.

The two-dimensional vertical frequency/temporal frequency filters required to implement the method can be realised by combining signals from different lines and fields in proportions related to the Fourier transform of the required passband defined in the two-dimensional frequency space.

Again it is seen in relation to this second embodiment that a two-dimensional high-pass filter is used to extract a high-pass sig nal of the spectral form of Figure 10 which can be sub-sampled vertically by a factor of two in relation to the source signal standard without interaction. Atwo-dimensional high-pass interpolator is used to provide a high-pass signal up-converted vertically by a factor of two from the spectrum of Figure 11 to the high-pass regions of Figure 15. By the interlaced sub-sampling of the high-pass signal the vestigial impairments of the reconstituted signal can be eliminated by suitable choice of two-dimensional filters.

In this embodiment, it has been assumed that the auxiliary detail signal has the same bandwidth as the main signal. That is, it does not enhance the horizontal resolution of the main signal, but only the vertical and temporal resolution. If the bandwidth of the main signal is enhanced it may even be possible to reduce the bandwidth of the auxiliary signal. Then it would be possible to carry further information in the auxiliary signal on a subcarrier. This could include a low-bandwidth signal representing the point-by-point displacement of right and left images in a three-dimensional system Alternatively such information could be coded digitally in the line blanking interval.

Claims (17)

1. A method of transmitting an input video signal having a relatively high definition format over two channels, namely a main channel signal which is compatible with a standard format video signal, and an auxiliary channel, in which the high definition input signal is filtered with a low passband in relation to repetition frequencies of the signal to provide the compatible main channel signal, and the high definition input signal is filtered with a high passband in relation to repetition frequencies of the signal to provide the auxiliary channel signal.

2. A method according to claim 1, in which the high definition input signal has 4N+1 lines per picture with a 2:1 interlace, and the main channel comprises a main signal which is compatible with a standard format video signal having 2N+1 lines per picture with a 2:1 interlace.

3. A method according to claim 1 or 2, in which the auxiliary signal is transmitted in the same format as the main signal.

4. A method according to claim 1, 2 or 3, in which the lowpass filtered signal is subsampled to provide the main channel signal and the highpass filtered signal is subsampled to provide the auxiliary channel signal.

5. A method according to claim 4, in which the sub-sampling operations are such that sub-sampling positions for the auxiliary channel are interleaved in relation to the sub-sampling positions for the main channel.

6. A method according to claim 4 or 5, in which the sub-sampling is applied to a filtered signal having sample positions which include the repetition positions at the said repetition frequencies of the input video signal as well as the sub-sample positions of the compatible main output signal.

7. A method according to claim 6, in which the said filtered signal is derived by adding notional zero-value positions to the input signal prior to filtering.

8. A method according to any of claims 1 to 7, in which the low and high passbands are defined in terms of the horizontal sampling frequency of the video signal.

9. A method according to any of claims 1 to 7, in which the low and high passbands are defined in terms of the vertical frequency and temporal frequency characteristics of the signal.

10. Apparatus for use in the method of claim 1, comprising main channel separation means for deriving from a high definition signal a main channel signal which is compatible with a standard format video signal, and in which the high definition input signal is filtered with a low passband in relation to repetition frequencies of the signal to provide the compatible main channel signal, and auxiliary channel separation means for deriving an auxiliary signal from the high definition signal, and in which the high definition input signal is filtered with a high passband in relation to repetition frequencies of the signal to provide the auxiliary channel signal.

11. A method of regenerating a high definition video signal from main and auxiliary signals produced by the method of claim 1, comprising filtering the main signal to provide interpolated lowpass filtered signals in relation to the repetition frequencies of the signal and filtering the auxiliary signal to provide interpolated highpass filtered signals in relation to the repetition frequencies, and combining the two resultant signals.

12. Apparatus for regenerating a high definition video signal from main and auxiliary signals produced by the method of claim 1, comprising main channel interpolation means for providing interpolated lowpass filtered signals in relation to the repetition frequencies of the signal, auxiliary channel interpolation means for providing interpolated highpass filtered signals in relation to the repetition frequencies of the signal, and means for combining the two resultant signals.

13. A video signal transmitter substantially as herein described with reference to Figure 2.

14. A video signal receiver substantially as herein described with reference to Figure 2.

15. A video signal transmitter substantially as herein described with reference to Figure 16.

16. A video signal receiver substantially as herein described with reference to Figure

17.