Patent Application
20140064379
This invention is related to video compression and decompression systems in which chrominance information is sub-sampled relative to luminance information and in particular to such systems where the chrominance sub-sampling varies between horizontal and vertical directions.
It is well understood that since the human visual system is much more sensitive to variations in brightness than colour, a video compression system need devote less bandwidth to chrominance information (typically colour difference components Cb and Cr) than to luminance information, the luminance component being usually denoted Y. Using the standard format notation in which 4:4:4 indicates no chrominance sub-sampling, video compression systems commonly utilise 4:2:0 in which Cb and Cr are each sub-sampled at a factor of 2 both horizontally and vertically. In for example H.262/MPEG2 or H.264/AVC, a macro-block may contain four 8×8 luminance blocks but only one Cb block and only one Cr block.
In 4:2:0, (as well as of course in 4:4:4) there is uniform sampling of chrominance in both horizontal and vertical directions, that is to say the chrominance information is sampled at the same sample densities in the horizontal and vertical directions. In high quality professional applications (for example CCIR 601) it has long been common to employ 4:2:2 in which Cb and Cr are each sub-sampled at a factor of 2 in only the horizontal direction. Thus, chrominance information is sampled at different sample densities in the horizontal and vertical directions respectively.
It is an object of this invention to provide more efficient techniques in encoding and decoding to accommodate video formats such as 4:2:2 where chrominance chrominance information is sampled at different sample densities in the horizontal and vertical directions respectively.
In one aspect, the present invention consists in a method of video encoding comprising the steps of receiving a video input in a first chrominance sampling format in which chrominance information is sampled at different sample densities in the horizontal and vertical directions respectively; resampling the video to a second chrominance sampling format in which the chrominance information is sampled at the same sample densities in the horizontal and vertical directions; forming residuals in which chrominance information is sampled in the second chrominance format through the use of reference samples in the first chrominance format; and transforming, quantising and entropy coding the residuals to form an encoded bitstream in the second chrominance sampling format. The encoded bitstream may contain a message indicating a chrominance resampling from the first chrominance sampling format.
Suitably, the encoder has a first mode of operation in which the second chrominance sampling format is up-sampled with respect to the first chrominance sampling format and a second mode of operation in which the second chrominance sampling format is down-sampled with respect to the first chrominance sampling format; and wherein a message in the encoded bitstream indicates the mode of operation.
The encoder may select between a plurality of chrominance resampling filters; wherein a message in the encoded bitstream indicates the selection of filter for the decoder
In another aspect, the present invention consists in a method of decoding an encoded bitstream comprising the steps of: receiving the encoded bitstream in a second chrominance sampling format in which the chrominance information is sampled at the same sample densities in the horizontal and vertical directions; performing inverse entropy coding, quantising and transforming steps to provide a residual in the second chrominance sampling format; reconstructing a video output from the residual and a predictor computed from decoded samples in a first chrominance format in which chrominance information is sampled at different sample densities in the horizontal and vertical directions respectively; and resampling to the first chrominance sampling format before or after said reconstruction.
Where the encoded bitstream contains a message indicating a chrominance resampling from the first chrominance sampling format; the step of receiving the encoded bitstream in the second chrominance sampling format may comprise decoding said message with the step of resampling to the first chrominance sampling format being conducted in response to said message. The resampling may switch in response to said message between down-sampling and up-sampling. The resampling may be conducted by a chrominance resampling filter selected in response to said message from a plurality of chrominance resampling filters.
Suitably, the step of resampling to the first chrominance sampling format is completed before said reconstruction. Alternatively, the step of resampling to the first chrominance sampling format is completed after said reconstruction and the predictor is resampled to the second chrominance sampling format.
In another aspects, the present invention consists in a video encoder configured to implement the above encoding method; a video decoder configured to implement the above decoding method; and a non-transitory computer program configured to cause programmable apparatus to implement either.
Video compression formats that have some components sub-sampled in only one direction (e.g. 4:1:1 or 4:2:2) are normally compressed in a way that such format is used as an input to the encoder and encoder's compression algorithms also operate on such sub-sampled chroma pixels. This invention proposes an alternative where, during encoding a video can be processed so that uniform sampling in both directions is achieved. To reconstruct the native format, during decoding an additional step is used to re-sample (down-sample or up-samples) decoded pixels. Such a video codec performs basic compression methods (transform, quantisation, entropy coding) on uniformly sampled signals. The benefits of such codec are:
Additional functionalities can be introduced which give more freedom for content adaptation during the compressed bitstream creation stage. These are introduced by for example allowing choice of down- or up-sampling filters which can be dynamically changed during compression. Decoding parameters can be adapted to provide desired decoding output.
Reference is directed to
Within the local decoder provided at the encoder, there is a decoder re-sampler (DR). So, after inverse quantisation Q−1 and inverse transformation T−2 and addition of the previous prediction, the locally decoded video is resampled to 4:2:2 format before storage in the reference frame buffer which provides the reference samples for the prediction. At some point in the prediction path between the reference frame buffer and the subtractor which forms the residual for transformation at T, resampling will be conducted to ensure that the prediction which is subtracted is in the same chrominance sampling format as the resampled video from which it is subtracted.
Resampling will typically be chrominance up-sampling to 4:4:4.
It may be helpful for the encoder to have an additional mode of operation in which chrominance is down-sampled to 4:2:0. It may also be helpful for the encoder, whether up-sampling or down-sampling to have a selection of resampling filters available to it. The encoder may then make decisions on whether to up-sample or down-sample and on which resampling filter to select, based on video content or other relevant parameters and constraints. The encoder will signal its decisions to the decoder so that the decoder may employ complementary resampling. Specifically, format parameters signalling the use of re-sampling and—where appropriate—conveying information about the type of re-sampling (up or down); the chrominance re-sampling filter used and the nature of the prediction loop are provided to the entropy encoder and bit-stream forming block for incorporation in the bit-stream as a message.
An example of an application scenario is given in Table A which represents a compression timeline of video in 4:2:2 format. The sampling filter required at the decoder is communicated in the compressed bit-stream. The filter notation is used:
Another simple scenario can also be achieved, as demonstrated in Table B. In this example, uniform sampling is achieved by repetition of pixels in the direction with lower sampling rate.
Although core compression is performed on uniformly sampled video, the decoder receives a signal indicating decoding to another format (e.g. 4:2:2). The decoder is then capable of re-sampling the signal. For example,
Possibilities for decoder design include:
While such re-sampling can happen after full reconstruction of uniformly sampled video, a decoder as shown
Decoded frames in native 4:2:2 format are used for forming the prediction. On the other hand, the residual coming from the transform T is uniformly sampled. Therefore the prediction has to be resampled to uniform sampling. After this step of reconstruction, in which the prediction is added to the residual, the reconstructed frame is in uniformly sampled format. Before any full frame filtering and outputting of decoded video, the chroma samples are converted to the native 4:2:2 format.
An advantage of the arrangement shown in
It will be understood that this invention has been described by way of example only and a wide variety of modifications are possible without departing form the scope of the invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 1215940.6 | Sep 2012 | GB | national |
