Polarization restraint based fast motion estimation approach to stereoscopic video coding

An efficient polarization restraint based fast motion estimation algorithm for stereoscopic video coding was proposed after analyzing polarization restraint principle in a parallel binocular-camera system. The left-and right-view images were compressed as the base and enhancement layer respectively in the proposed scheme. An optimal motion vector predictor for the right channel was attained after combining disparity estimation and motion estimation (ME), and a final motion vector was got by using different motion searching method. If the final motion vector predictor was got by referring to that of left channel, then ME searching only in horizontal direction within a small range was sufficient according to polarization restraint, and thus the ME complexity in the right view channel could be greatly reduced. Experimental results show that the ME time of proposed algorithm is just 0.27, 0.58 and 0.67 times of that of full search block matching algorithm, diamond search and three step search respectively, accelerating coding speed greatly.