Optical flow estimation method to determine compensation by multi resolution of hierarchical structure

Abstract

Optical flow is appearance of the object motion in image represented by the velocity vector distribution. In this research we use a spatiotemporal differentiation method to obtain optical flow. This method enables us to calculate velocity distribution rapidly, but the error in the approximation of derivative coefficients increases as the displacement of the moving pattern between the successive frames becomes large. This is a crucial principle defect of the spatiotemporal differentiation method. To expand the range of measurable velocity using spatiotemporal differentiation, we use the compensation method. In previous research, we used the method to divide the image into some local regions, and decided the amount of compensation based on the variance of estimated velocity. In the compensation method, it is important to decide the amount of the optimum compensation. In this research, we introduce how to decide the amount of compensation in view of the image of multi-resolution. Experimental results show that more accurate velocity is attained.