Motion Analysis Based on Invertible Rapid Transform

Abstract

This paper presents the results of a study on the use of invertible rapid transform (IRT) for the motion estimation in a sequence of images. Motion estimation algorithms based on the analysis of the matrix of states (produced in the IRT calculation) are described. The new method was used experimentally to estimate crowd and traffic motion from the image data sequences captured at railway stations and at high ways in large cities. The motion vectors may be used to devise a polar plot (showing velocity magnitude and direction) for moving objects where the dominant motion tendency can be seen. The experimental results of comparison of the new motion estimation methods with other well known block matching methods (full search, 2D-log, method based on conventional (cross correlation (CC) function or phase correlation (PC) function) for application of crowd motion estimation are also presented.