Matching of Affine Transformed Images by Using Similarity Based on Local Concentric Features
This paper proposes a pattern matching method based on concentric features calculated over local areas close to feature points. For such feature points, the corner points of images are used, and images generated by complex-log mapping and Fourier transform are used as the concentric features. The procedures are as follows: (1) Similarity is found based on the concentric features in neighborhoods of corner points; (2) in consideration of the uniqueness of correspondence and removal of pseudo-correspondence, correspondence is obtained from this similarity; (3) with correspondence as a weight, the parameters of the affine transformation are estimated. By conducting experiments, the robustness of the proposed technique against deformations and noises is shown.