Comparison of accumulative computation with traditional optical flow

Abstract

Segmentation from optical flow calculation is nowadays a well-known technique for further labeling and tracking of moving objects in video streams. A likely classification of algorithms to obtain optical flow based on the intensity of the pixels in an image is in (a) differential or gradient-based methods and (b) block correlation or block matching methods. In this article, we are going to carry out a qualitative comparison of three well-known algorithms (two differential ones and a correlation one). We will do so by means of the optical flow obtaining method based on accumulated image differences known as accumulative computation.