A Motion Vector Coding Scheme Based on Bottom-Up Merging Procedure

This paper presents a new motion vector coding scheme based on bottom-up merging procedure. For low motion video sequences, most of the storage space of the motion vector data spending on the low motion objects or stills back ground. The proposed scheme merges the neighboring macro-blocks with similar motion vectors into bigger rectangular blocks, and those merged blocks may represent as all or a part of the objects and the background. For the merged blocks, each block will need only one motion vector to reconstruct it. Experimental results show that the proposed scheme can reduce about 50% motion vector coding bits as compared to the conventional way of storing motion vectors for each macro-block.