GB2351865A - Video compression - Google Patents

Abstract

An MPEG-2 encoder has a dumb coder which can slave to external coding decisions. A separate front end (which may be a standard MPEG-2 encoder) also operates on the incoming video to provide coding decisions for the dumb coder in the absence of external coding decisions.

Description

2351865 VIDEO COMPRESSION This invention relates to video compression,

especially the MPEG-2 standard, though it can be applied to any video compression system that is liable to degradation when coding and decoding are cascaded.

The concept of the Information Bus has already been introduced in an earlier patent application WO 95135628. This is a signal which carries information about an MPEG bitstream for use in a downstream process, for example, the re-encoding of a decoded MPEG picture. In a development of the Information Bus, known as the Information Stream [see WO 98103017], the Information Bus signal is formatted in such a way that it forms part of a decoded digital video signal and can thereby pass through standard studio equipment.

An MPEG encoder which receives the Information Stream within the incoming video signal, is not required to take coding decisions and can be a "dumb" coder. In many circumstances, however, the same encoder is required also to encode video signals which do not include the Information Stream. The encoder is therefore provided with a front end which can, in the absence of a received Information Stream, supply coding decisions to the dumb coder.

The MPEG-2 standard is in widespread use and a variety of integrated circuits for encoding, are commercially available. These commercially available encoding units are of course considerably more economic than circuits designed for special purposes.

It is an object of the present invention to provide an improved coder which can tale advantage of compression coding decisions, whenever these are carded in the input video, but which can utilise certain conventional (and thus more economic) circuit elements.

Accordingly, the present invention consists in one aspect in a video compression encoder, comprising an video input; means for removing compression coding decisions carried in the input video; and a dumb coder which receives the video signal and the compression coding decisions and is adapted to compression encode the video in accordance with the codling' decisions, characterised by the provision of a internal encoder operating on the input video to produce an internal coded signal and means for supplying the coding decisions taken in said internal coding to said dumb encoder for use in the absence of compression coding decisions carried in the input 5 video.

It will be understood that the internal coder may take the form of a conventional MPEG encoder and may therefore be constructed from commercially available integrated circuits. There is clearly some overlap in functionality between the internal coder which is here used to derive coding decisions and the dumb coder which - in the absence of an Information Stream in the input video - operates on those coding decisions. However, the larger part of the cost and complexity in the conventional coder lies in functions such as mode selection and motion vector generation,which are of course not required in the dumb coder. The saving in using a conventional encoder, rather than a dedicated circuit, to perform mode selection and motion vector generation, therefore far outweighs the apparent redundancy.

If the appropriate outputs are available, the internally taken coding decisions can be supplied directly to the dumb coder. Alternatively, the internal coding decisions can be inferred from analysis of the internal coded signal; this analysis being suitably performed by an Information Stream generating decoder as disclosed in the above references.

In another aspect, the present invention consists in an architecture for compression re-encoding based on the Information Stream, characterized in that the video input is encoded using a standard compression encoder, ignoring the Information Stream; the resulting compressed bitstream is decoded by an Information Stream generating decoder; the resulting second Information Stream, together with the original video input including the original Information Stream, are fed to a simplified Information Streambased reencoder, the simplified Information Stream-based re-encoder consists of a video and Information Stream compensating delay, priority selection means for selecting between the delayed original Information Stream and the second Information Stream to produce a third Information Stream, an Information 3 Stream Decoder and a Dumb Coder which encodes the delayed input video signal in accordance with the decoded third Information Stream The invention will now be described by way of example with reference to the accompanying drawings, in which: 5 Figure 1 is a block diagram of a cascaded decoder and re-encoder according to the prior art; and

Figure 2 is a block diagram of an encoder according to the present invention.

A block diagram of a cascaded decoder and re-encoder using the Information Stream is given in Figure 1. Systems according to Figure 1 have been shown to be successful. A drawback with the current system is that the architecture of the re-encoder side (everything to the right of the 'studio equipment' in Figure 1) is not suited to the use of standard MPEG2 components. Readily available MPEG-2 coding chipsets perform the two functions 'MPEG coder front end' and 'MPEG Dumb Coder' but do not make available a suitable interface between those two blocks for use as shown. In order to implement an encoder based on the Information Stream, it has therefore hitherto been necessary to make custom versions of both blocks. In particular, the 'MPEG coder front end'which includes motion estimation and prediction mode selection, is highly complex.

The architecture provided by the present invention overcomes the main problem of the prior art which is the need for a custom-designed coder front end including motion estimation and prediction mode selection functions with appropriate interfaces to a Dumb Coder. It enables the motion estimation and prediction mode selection functions of standard, low-cost encoders to be used while still retaining the benefits of Information Stream- based compression re-encoding.

Referring to Figure 2, the input to the re-encoder is a digital video signal Video I and its associated Information Stream signal Information 4 Stream 1. The Video I signal is encoded using a standard MPEG encoder to produce an MPEG bitstream which is subsequently decoded by an Information Stream generating MPEG decoder to produce Video 2 and Information Stream 2 signals. The original input signals are delayed in a compensating delay, which is controlled either by Video 2 or by Information Stream 2 to ensure that its Delayed Information Stream I output is co-timed with Information Stream 2.

Delayed Information Stream 1 contains the re-encoding information present in the original input, which may be invalid for some pictures or macroblocks because of upstream processing operations. Information Stream 2 contains equivalent re-encoding information which is valid at all times but which need only be used whenever Delayed Information Stream 1 is invalid. The Information Stream Switch is therefore used to select between its two inputs, choosing Delayed Information Stream I if it is valid, otherwise choosing Information Stream 2. The resulting Information Stream 3 is then decoded to Information Bus format and used to control a Dumb Coder working on the suitably timed Delayed Video 1 signal.

The advantage of this architecture is that the only encoding decisions that need to be made by the Dumb Coder are quantizer decisions made by the bit-rate control process, together with possible changes of macroblock type or picture type. These latter decisions could mostly be avoided if the standard MPEG decoder operating on the input video signal were able to base some of its picture-rate encoding decisions (such as GOP structure) on the input Information Stream, as indicated by the dotted line in Figure 2. in that case, an additional simple processing block may be required at the input to the standard encoder to convert the Information Stream format into one which can be read by the encoder. In any case, the need for custom circuits to implement the most computationally intensive encoding decisions, including motion estimation and prediction mode selection, has been avoided in this architecture.

Note that the Information Bus and Information Stream are simply two different formats for the same data, so the exact configuration of blocks in Figure 2 could be replaced, for example, with an arrangement in which the Information Stream signals are decoded to Information Bus format before being switched.

If the standard MPEG encoder makes these available, appropriately 5 formatted coding decisions could be supplied directly to the Information Stream switch.

6

Claims (5)

1. A video compression encoder, comprising an video input; means for removing compression coding decisions carried in the input video; and a dumb coder which receives the video signal and the compression coding decisions and is adapted to compression encode the video in accordance with the coding decisions, characterised by the provision of a internal encoder operating on the input video to produce an internal coded signal and means for supplying the coding decisions taken in said internal coding to said dumb encoder for use in the absence of compression coding decisions carried in the input video.

2. An encoder according to Claim 1, in which the internal coder takes the form of a conventional MPEG encoder.

3. An encoder according to Claim 1, wherein the internal coding decisions are inferred from analysis of the internal coded signal.

4. An encoder according to Claim 3, wherein said analysis is performed by an Information Stream generating decoder.

5. An architecture for compression re-encoding based on the Information Stream, characterized in that the video input is encoded using a standard compression encoder, ignoring the Information Stream; the resulting compressed bitstream is decoded by an Information Stream generating decoder; the resulting second Information Stream, together with the original video input including the original Information Stream, are fed to a simplified Information Stream-based re-encoder, the simplified Information Stream-based re-encoder consists of a video and Information Stream compensating delay, priority selection means for selecting between the delayed original Information Stream and the second Information Stream to produce a third Information Stream, an Information Stream Decoder and a 7 Dumb Coder which encodes the delayed input video signal in accordance with the decoded third Information Stream.