Image Registration by Contour Matching Using Tangent Angle Histogram

Image registration is a fundamental task in image processing. In this paper, a novel contour-based image registration approach is proposed for images with affine distortion. In the proposed algorithm, the normalized histogram formed by the tangent angle of the contour is constituted. The peak in the circular correlation result of these two histograms indicates the rotational difference between images. Other affine parameters are determined using the ratio of length of contour and the position translation of centroids, thus the coarse registration result is determined. After refinement, the final registration result is acquired. The contour matching algorithm proposed in the paper can be used for both open and closed contour. Experimental results demonstrated that this approach is an effective way to register images automatically.