Methods for the Synchronisation of
Successive Digital Video
Compression/Decompression Systems
The present invention relates to the processing of digital signals and more particularly to digital video signals.
Digital video compression algorithms such as those disclosed in ISO/IEC 11172-2 (MPEGl) and 13818-2 (MPEG2) have already been proposed and are used to reduce the bandwidth necessary for the transmission of video signals in digitally coded format. In order to reduce the temporal redundancy in video sequences such coding schemes usually predict future pictures (e.g. digitally coded television frames) from previously decoded ones. By predicting pictures more than one frame into the future, it is possible to code intermediate frames by predictive processes from past as well as future frames as outlined in the standards MPEGl and MPEG2. This predictive coding process is then, often at regular intervals, replaced by a complete reconstruction of the current frame (intra-frame coding) in order to constrain the propagation and accumulation of coding errors and to provide unambiguous start points for the decoding process.
If two or more of such digital encode/decode processes are cascaded it is important that the integrity of intra-coded frames is preserved. This means that pictures which have been intra-coded in the first compression process are again intra-coded in all compression processes that are subsequently applied to the same video sequence. If this constraint is applied, error accumulation is prevented over the entire system.
It is important to note that, as digital video compression systems become more common-place, the
requirement outlined above becomes more and more difficult to achieve since successive coding systems might be at physically different locations. More importantly, a video signal which has undergone digital compression and decompression might be stored and edited in decompressed form (e.g. on an analogue video cassette recorder) and be subsequently re-encoded for transmission purposes. Figure 1 shows an example of a cascade of two successive compression-decompression systems currently being used, with the intermediate signal (c) stored, for example, on a video cassette recorder. This process of compression/decompression storage and/or transmission could be repeated several times. If the compression systems in such a cascade of successive coding schemes operate entirely independently of one another the picture quality rapidly deteriorates. It is, therefore, important that, if transcoded frames are used for refresh in such a cascade the same frames are intra-coded throughout the entire chain of encoders and decoders.
The present invention proposes a method which makes it possible to synchronise cascaded compression systems in such a way that frames which have been intra- coded in the first coding system will then be intra- coded by the second and, therefore, all following coding systems.
According to the present invention, a method is provided for conveying information in a video signal, comprising the steps of adding code signals to the video signal at certain points in a succession of video signals representing a frame of video, said code signals relating to the process used to create said frames, and using said code signals for choosing a coding
scheme for said signal coding said video signal using said coding scheme, decoding said video signal such that the resulting decoded video signal comprises said code signals.
In order that the present invention be more readily understood an embodiment will now be described by way of example with reference to the accompanying drawings in which:
Figure 1 shows an example of a cascade of two successive compression systems with intermediate storage device.
Figure 2 shows a block diagram of the proposed method using a cascade of two successive compression/ decompression systems.
Figure 3 shows a block diagram of proposed method B, with detector and inserter prior to the encoder.
Figure 4 shows the temporal alignment of signalling waveform with coding blocks.
A first embodiment of the invention, is shown in Figure 2, which shows a cascade of two successive compression systems, each consisting of an encoder and a decoder.
A signal to be processed (a) is fed to an encoder (1) and the encoded signal (b) is then fed to a decoder (2) of the first system. The decoder (2) of the first system sends the coding information (d) of the first system to the signal inserter (3). Signal inserter (3) codes the information on the video signal (e) as described below. This information is then conveyed together with the video signal (e) to the signal detector (4) of the second compression system. Signal detector (4) extracts the information from the video signal (e) and passes this information on to the
encoder (6) of the second compression system. Since signal inserters are used at the output of all decoders, several of such systems can be cascaded. In this simple mode of operation the blanking unit (5) simply passes the video signal on to the second encoder (6).
In its simplest form, the signalling information is coded in the vertical blanking interval at the top of the frame to which the signalling information applies. An example of such an application is the use of a user data field as defined in ITU-R Rec. 656 digital video standards for the transmission of the intra-frame signalling bit. In cases where the transmission medium between inserter (3) and detector (4) can be guaranteed to be transparent to the vertical blanking interval this simple method is all that is needed to synchronise successive video compression systems.
The disadvantage of the embodiment described above is that it relies on the transmission/ storage medium being transparent to the vertical blanking interval. If such transparency cannot be guaranteed the signalling information should be coded in the active part of the video signal itself as described below.
The signal waveform is chosen such that it survives most commonly used analogue and digital recording and transmission formats under worst-case broadcast quality signal-to-noise conditions. It could consist, for example, of a series of luminance pulses, such as raised-cosine pulses, with a width of 0.3 us or more. These could be located at the beginning of the last active line of the video frame prior to the frame to which the signalling information applies. This line is normally hidden in the over-scan region of display monitors. By putting the information into the last line of the previous frame the second encoder (6) has enough time to synchronise its intra-coding frames to that of
the previous coding system.
In order to synchronise successive video compression systems with respect to the location of intra-coded refresh frames, only a single bit of information needs to be conveyed from the decoder (2) of the first system to the encoder (6) of the second. If this is done with a single pulse as described above, the signal detector (4) has to be disabled manually in cases where signalling information is not available or when the signal-to-noise ratio is too low for reliable detection of the signalling bit. Otherwise the encoder (6) would try and synchronise its intra-frames to spurious video waveforms or random noise which would be highly undesirable.
For the proposed method to be of any use in a real application the signal detector (4) needs to detect not only the signalling bit itself but it must also make a decision as to whether signalling information is present at all and, if it is, whether it is of sufficient quality in terms of its error rate to be used for controlling the encoder (6).
This can be achieved by coding synchronisation words instead of signal bits. Since the information is repeated over many frames (there are usually about 10 predictively coded frames for every intra-coded one) extension to synchronisation words of only a few bits provides enough redundancy to make reliable detection of the signalling information possible.
A further advantage of automatic signalling detection is the possibility of removing the signalling information in the blanking unit (5) prior to re- encoding in (6). This structure makes it possible to hide the signalling waveform from the viewer even in cases where the display monitor is not over-scanning the picture.
The disadvantage of the embodiment as
described above is that it needs additional circuitry at the decoder as well as the encoder. This problem can be avoided by moving both the detector and the signal inserter in front of the encoder, thus avoiding any additional circuitry at the decoder. Figure 3 shows a proposed third embodiment in block diagrammatic form.
If the detector (2) detects the presence of a signalling waveform, it passes the information (c) both to the encoder and to the inserter (1). If no signal is detected, the coding information (d) is entirely determined by the encoder (3) .
This method places additional requirements on the signalling waveform in so far as it now has to survive not only analogue transmission and storage formats but also digital compression. To achieve this, the proposed signal consists of a series of raised- cosine pulses, each, for example, 32 pixels wide. The centre pixels of each symbol (e.g. 16 pixels) are kept at a constant level and are aligned with the coding blocks of the compression system, thus minimising the distortion of the signalling waveform due to compression. Figure 4 shows how the signalling waveform is aligned with respect to coding blocks of the compression system.
The essential feature of the signalling waveform is that it is kept at a constant value for the duration of one or more complete compression blocks so as to minimise the distortion compression would have on the waveform. By inserting guard intervals between successive bit levels cross-talk between symbols is minimised. Since all symbols have to be aligned with the coding structure, the guard intervals must have a minimum period equal to the horizontal size of at least one coding block. The waveform between symbols, however, is less important but would normally be chosen to have low spectral content.
Further to the two modes of compressing video
sequences described above (intra-frame coding and forward predictive coding) coding schemes such as those described in ISO/IEC 11172-1, MPEG-1, Coding of Moving Pictures and Associated Audio, Part 2, Video and ISO/IEC 13818-2, MPEG-2, Coding of Moving Pictures and Associated Audio, Part 2, Video, allow the definition of bi-directionally predicted pictures. These are pictures which are predicted both from a past and a future reference picture. In order to achieve the best picture quality after several cycles of compression and decompression in systems using bi-directional prediction it may be desirable that the entire sequence of intra, bi-directional and forward predictive coding is faithfully repeated in all coding/decoding cycles.
Such a system necessitates the signalling of three discrete values. The relative importance of these values in terms of the effect of signalling errors on the overall picture quality is, however, by no means uniform. Cleariy, the alignment of intra-coded frames is much more important than the distinction between forward and bi-directional prediction. The proposed method can take this non-uniformity into account by putting additional error protection or the information about intra-coding whereas other coding parameters might need less error protection.
Further examples of additional coding parameters which might be signalled from one compression system to the next are the use of field or frame coding, the type of block scanning pattern and the quantisation table used in the first encoding process.
From the above description it will be seen that in systems where a video signal undergoes several stages of digital compression and decompression it is highly desirable that frames which have been intra-coded in the first stage are also intra-coded in all
subsequent stages of the system. The proposed method makes it possible to achieve such synchronisation with a minimum of circuitry.
In particular, it is shown how a signalling waveform can be detected and inserted prior to the first stage of a compression encoder thus avoiding any additional circuitry in the decoder. While the method is most suitable to synchronising successive compression systems, its application can be extended to any kind of information conveyed in the active part of the video signal.