An efficient matching approach to motion analysis of images

Local information in not always enough for efficient motion analysis. Additive processing is required to get accurate results. This has been formulated as the aperture problem. Block matching algorithms are applied between successive images for motion estimation, assuming conservation of local intensity distribution. Matching approaches provide good results when the aperture problem does not exist. However, in regions when the aperture problem exists, additional constraints are required in order to recover the displacement of pixels between consecutive images. In this paper we present a way to improve the performance of optical flow computation at the first early level. Morphological filters are introduced in the matching approach with which we overcome inherent problems of correlation based techniques.