A Low-Complexity, Sequential Video Compression Scheme Using Frame Differential Directional Filter Bank Decomposition in CIE Lab Color Space

This paper proposes a novel differential directional filter bank (DFB) decomposition-based video coding scheme. DFB decomposition is generally preceded by Laplacian pyramid decomposition, as in contourlet transform, in order to separate the high pass content from an image or a video frame. This leads to redundant representation in the transform domain. In the proposed work, increase in redundancy due to contourlet transform is compensated by the application of DFB decomposition directly to the difference between motion compensated consecutive frames, called residual frames, which capture temporal discontinuities in a video. Low-complexity adaptive rood pattern search scheme is used for motion-estimation and tiled version of set partitioning embedded block coder is proposed which reduces the encoding time. The proposed compression scheme is implemented in CIE La*b* color space to exploit its advantages like; linearity and perceptual uniformity. The proposed low complexity sequential video coding scheme achieves, on an average, 0.71 dB more PSNR than H.264 while requiring up to 17.6% less encoding time compared with the H.264 baseline encoder on standard test video sequences.