Unit gradient vectors based motion estimation techniques

This paper presents two motion estimation techniques based on gradient orientation information (GOI). We utilize GOI by means of unit gradient vectors (UGVs) of an image. UGVs are used, in place of image intensities, in two conventional motion estimation techniques: the gradient method (GM) and the gradient structure tensor method (GSTM). We name the two novel approaches the gradient orientation based gradient method (GOGM) and gradient orientation structure tensor method (GOSTM), respectively. The proposed methods have been compared with the traditional methods, the GM and GSTM, under mildly noisy (40dB) and uniformly varying lighting conditions. As the gradient orientation is remarkably invariant to uniform changes of image intensities, simulation results show that the proposed methods can perform motion estimation robustly in changing lighting conditions. The results also reveal the signiflcance of a low-pass fllter to be applied to an input image sequence prior to the computation of motion vectors. The paper also discusses the appropriate size of the local area where a motion vector is estimated.