MOTION ESTIMATION FOR SCALABLE VIDEO CODING

Abstract

Motion estimation is important component of video coding systems because it enables us to exploit the temporal redundancy in the video sequence. Generally, block based motion estimation algorithms are used which search the macro-blocks in the reference frames for the best match in the vicinity of the location of current macro-block. Estimation can .further be improved by searching for matching macro-blocks at sub-pixel positions in the reference frame. Motion estimation is a computationally expensive process and the complexity increases as the sub-pixel resolution of motion vectors is increased. For encoding spatially scalable video, motion estimation process is required to be performed for each of spatial resolution. Therefore, such techniques are required for motion estimation for scalable video coding which are less computationally expensive and also provide good PSNR performance. Activity based motion estimation technique which has been proposed recently, dynamically adapts the search range of motion estimation in enhancement layer based upon the activity of corresponding macro-block in base layers. In this thesis we have proposed a new motion estimation scheme which decreases the overall complexity of activity based motion estimation scheme while maintaining similar PSNR performance.