Optimized image registration method integrating color histogram and cosine similarity

The invention discloses an optimized image registration method integrating color histogram and cosine similarity. The method comprises an image preprocessing stage, an image feature bidirectional matching stage, a homography matrix calculation stage and an overlapping region determination stage. In the image preprocessing stage, an image block pair with the maximum similarity measurement value isobtained by using a blocking method and a color histogram and cosine similarity combined method; in the image matching stage, for the problem that the detected image features have wrong matching, a bidirectional matching method is adopted to eliminate wrong matching pairs; and finally, a homography matrix is calculated by using an RANSAC algorithm, and an overlapping region of a to-be-registered image is determined, and is taken as input of subsequent image fusion. The method is of great significance to acceleration of image registration of a large field of view, and a new solution is providedfor image registration. 本