Segmentation of motion field for scalable video compression

In this article we present new proposition of video compression based on a hybrid codec with segmentation. We proposed video segmentation algorithm based on block matching algorithm and Marcov Random Fields. This system works in two steps. In the first step motion vectors will be calculated. In the next step motion vector will be used in segmentation process. The second stage is an image classification based on motion vector and intensity function as well as compression with different rates of all segmentation regions from image. In this work we can also find influence of different macroblocks shape for: estimation vector field, segmentation process and compression rate.