Block Matching Algorithm for Video Coding

Motion estimation is a key issue in the field of moving images analysis. In the framework of video coding, it is combined with motion compensation in order to exploit the spatio temporal correlation of image sequences along the motion trajectory. It then achieves one of the most important compression factors of a video coder. By dividing each frame into rectangular blocks, motion vectors are obtained via the block matching algorithms (BMA). The full search algorithm (FS) is a brute force BMA. It searches all possible locations inside the search window in the reference frame to provide an optimal solution. However, its high computational complexity makes it often not suitable for real-time implementation. Many fast but sub-optimal algorithms are introduced to improve the performance of video coders. The present book analyses three prospects of improving the quality of existing video coding schemes. Namely, one at a time optimization, adaptive search stagey and feature domain based criteria.