Motion Compensation Algorithm Based on Color Orientation Codes and Covariance Matching

Regions extraction and matching are two key steps in image motion compensation. This paper applies the skills of covariance matching and information entropy to motion compensation to improve its performance. First, color orientation codes are employed to compute the information entropy of an image region and detect the sub-block regions automatically. Secondly, covariance matrices are used to match sub-blocks between current frame and previous frame. Finally, we gain the global motion parameters by affine motion model. Experimental results show that the proposed algorithm can detect and compensate global motion in indoor and outdoor environment and has outstanding result than traditional histogram matching.