Efficient motion estimation algorithm based on structure segmentation and homogeneity test

This paper reports a novel motion estimation technique for image sequence coding based on structure segmentation and homogeneity test. In this method, the frames are first segmented into four macroblock types corresponding to stationary background, uncovered background (or new objects), moving objects and edges among them. Then the following motion estimation is only performed on the compensated macroblocks according to motion homogeneous objects, and the motion interaction among neighboring macroblocks is further used for reducing the search range. Preliminary results indicate that this technique not only reduces the estimation burden drastically but also improves SNR and subjective picture quality.