A novel energy function and its minimization for video object segmentation and compression

Traditional digital video compression techniques focus on image-block encoding. Conversely, the MPEG-4 standard specifies that a video should be composed of Video Object Planes, improving compression through spatial cohesion. AlPEG-4 video compression can incorporate two key observations: First. a block containing different textures needs to be split to enhance compression. Second, Motion Vectors are detected more accurately by grouping blocks into regions, allowing for more efficient predictive coding. We propose a novel discrete cosine transform energy function EDCT measuring block compression for image and video segmentation. By minimizing EDCT the best possible split of the block can be found. Consistent motion vectors can be obtained by using an innovative energy function in the spatio-temporal domain which measures block motion over multiple frames. Tests on images and video show promising results. where still image compression achieves approximately 15% improvement over JPEG and our video compression achieves similar improvement over MPEG-2.