Study On Remote Sensing Image Feature Registration Based On Improved Edge Detection Operator

Image registration is the key step of image analysis and processing, which is essential prerequisite of image comparison, data fusion, change analysis and target recognition. But since the multi-source remote sensing imagery can provide more abundant, integrated and precise information than single-source image, the multi-source remote sensing image registration has becoming more and more important in the remote sensing application.Taken the multi-source remote sensing image registration as the research aim, a feature matching algorithm is presented based on improved Canny edge detection operator and the combination and optimize to the traditional registration methods. The method improves the registration precise and automatic degree obviously. The main contents are shown as following.1. The basic theories of image registration, transform models, resampling, similarity measures are briefly introduced, and the basic matchimg methods and their performances are summarized.2. The theories, methods and characteristics of general edge detection operators are analyzed and summarized systematically. Compared the experiment results of the existing edge detection operators, it is proved that the performance of classical Canny operator is optimal in all of the operators, but this classical Canny operator has its shortage.3. An improved Canny operator and the method to select threshold adaptively are recommended. These methods improved the orientation capability of edge detection and shorten the time of edge extraction.4. The experiments show that the feature matching algorithm based on the above improved operator increases the registration precise and automatic degree obviously.