Video Compression Scheme Using DEMD Based Texture Synthesis

In this paper we present a video coding scheme based on texture synthesis through Directional Empirical Mode Decomposition (DEMD). In this scheme P and B-frames of the video sequence are decomposed and parametrically coded with the help of DEMD algorithm, while I-frames are coded with the help of H.264. All P and B frames are decomposed into Intrinsic Mode Function (IMF) image and its residue. Only the first level IMF image for P and B frames are coded. At decoder stage subsequent IMF images are synthesized with the help of correlation search. Wavelet decomposition is performed over residual image and energy level at the HH band is used as a decision criterion for number of decomposition to be performed for optimum synthesis. The experimental result demonstrates the effectiveness of the algorithm in multi-resolution parametric modeling of image data which can be efficiently coded to achieve significant compression with acceptable quality. This scheme also enables to perform scalable coding of IMF parameters to achieve higher compression with perceptual quality.