Automatic Image Registration Based on Shape Features and Multi-scale Image Segmentation

Image registration plays an important role in many practical problems in diverse fields such as medical, remote sensing, and computer vision. The basic purpose of image registration is to obtain the finest geometrical and radio-metrically aligned image from temporal or multi-modal image sensors. In this paper, a novel stable shape feature-based image registration method has been proposed by matching the stable region with a set of rotations, scale invariant features, and multi-scale image segmentation is used to obtain the matching areas. This algorithm firstly converts images into image objects by multi-scale segmentation and convexity model restriction. Then these reliable and stable image regions are used as matching unit rather than points or lines. Experiments show the algorithm proposed in this paper is not sensitive to rotation and resolution distortion, which can accomplish image registration automatically.