Recovering translational motion parameters from image sequences using Randomized Hough Transform

Abstract

An algorithm for recovering of the direction of the translational motion from consecutive pairs of perspective images is proposed. The algorithm is based on the ideas of Randomized Hough Transform i.e. the principles of random sampling and accumulation of motion parameters. The translational motion parameters are solved with least-square approach from equation which relate 2-D motion field in image plane and 3-D structure and motion together. Some experiments based on simulated and real data are presented and they show that a robust interpretation of the translational motion can be obtained even in the presence of significant level of noise.