Horn & Schunck parameter characterization of the gradient based method for Optical Flow computing

Optical Flow is an algorithmic tool used in motion vision task in order to get the movement 3D and its projection 2D in image sequences. Today a great number of approaches have been developed and others continue appear. However, in many applications the algorithm of Horn & Schunck continue being used due to their simplicity and parallel capacity. Evidence of this is the steady progress by the increasing speed that presents the more recent works over FPGA platforms. However, despite of the speed has been increased, don't exist formal studies that aboard the parameter involved in the approach, and only in few cases has been addressed the smoothest term, the number of iterations and the accuracy at the moment that you are development an application based on the algorithm. The goal of this work is to characterize these values since the point of view of a set of image sequences and experimental processing. A set of image sequences contained in databases are analyzed as benchmark for images with the same nature. Experimental results are validated using the Angular and EndPoint Error whit the software MATLAB.