A multi-sensor image registration approach based on edge-correlation

In this paper we propose a multi-sensor image registration approach based on edge-correlation. Firstly, edges of images are extracted and the longer edges are coded by modified Freeman. Secondly initializing correlation of chain code is calculated by integrating correlation coefficient and other methods, then the consistency detection succeeds by relative distance ratio histogram clustering detector based on center of edge correlation and angle histogram clustering detector based on average directional angle difference of edge correlation. During the calculation of average directional angle difference of edge correlation, modified histogram approach in this paper is much better than line-fitting approach. Finally accurate edge correlation pair is obtained, and then image registration parameter is attained by applying least square algorithm (LSM) in interrelated parts. The approach presented in this paper can register images of wide translational and rotary range.