Line Correspondences Between Two Images Using Local Affine Moment Invariant

Abstract

This paper proposes an algorithm for matching line segments between two images which are related by affine transformations using local affine moment invariant (AMI). Instead of using traditional methods for objection recognition in which each object is globally represented by a vector of affine moment invariants, here each pair of line segments extracted from each image is locally represented by an affine moment invariant. This algorithm is suitable for line correspondences with multi-planes and occlusion. Matches are determined through comparing invariant values and voting. Experimental results are given for both synthetic and real images. The noise model of affine moment invariant is also presented.