A contex-based predictive coder for lossless and near-lossless compression of video

We propose a new approach to context-based predictive coding of video, where the interframe or intraframe coding mode is adaptively selected on a pixel basis. We perform the coding mode selection using only the previously reconstructed samples which are also available at the decoder, so that any overhead information on the coding mode selection does not need to be transmitted to the decoder. The proposed coder also provides the lossless concatenated coding property when applied to multigeneration of video sequences since the same coding mode information is available at the second time encoding. The proposed coding mode selection enables the coder to easily incorporate error modeling and context modeling by performing the intraframe coding with one of the existing image coders such as the JPEG-LS standard. Experiments show that the proposed approach in conjunction with the JPEG-LS standard provides significant improvements in compression efficiency.