A Fast Algorithm for Motion and Disparity Estimation Based on the Reference Directions of Neighboring Macroblocks

As for JMVC (Joint Multi-view Video Coding), when executing motion and disparity estimation simultaneously, the implicit correlation of reference directions between current and neighboring macroblocks is neglected, and this will lead to useless searching. To reduce this redundancy, we propose an estimation algorithm which sets search ranges properly by exploiting the mentioned correlation. The proposed algorithm first gets the reference directions of macroblocks in the left, top and top right direction of current macroblock, and compares them with current reference direction, then sets search ranges by the comparison results, executes motion and disparity estimation in the new ranges at last. Experimental results show that the proposed algorithm can save 28.98%- 46.30% of coding time without degradation of coding quality comparing to JMVC.