Abstract
A new efficient method of determining the amount of image change from a given pair of images using one dimensional section matching is presented. It is assumed that there are two long line segments with different orientations (characteristic lines) in these images. The image change parameters (the amount of translation ΔX and ΔY and the amount of rotation ψ) are calculated from the displacement vector of the intersection point of the two characteristic lines. The positions of the characteristic lines are determined using one dimensional image section matching. The intensity patterns of the two given images at a corresponding section are compaired and the characteristic line position is determined as the least sum-squared error point. The proposed algorithm includes only four one dimensional matching steps and is adequate for real time purpose. Experimental results are also presented (outdoor scene and laboratory scene), compairing with the results obtained by the point matching method (by identifying 5 corresponding point pairs in the two images and by solving a set of linear equations).
