Optical Flow Estimation Based on Predictive Vectors

In this paper, we analyze the problem of block matching method for computing optical flow and propose a new computing method, which records matching rate distribution for each block in the global searching area during the first time block matching procedure, and compute the predictive vectors, and then combining both of data information, compensate for the incorrect motion vectors, finally, compute correctly the global optical flow. Experimental results indicate that the approach is valid and robust.