A New Method in Image Matching Based on Spatial Relationships in Multi-Sensor Remote Sensing Images

Image registration process is one of the most important branches in the field of image processing, which is an essential preprocessing for the use of remote sensing. Scale invariant feature transform (SIFT) is one of the most commonly used feature-based methods for registration of images. However, a main weakness of this algorithm is the creation of a large number of mismatches. Based on the spatial relationships of the corresponding points of SIFT, the proposed method in this paper increases the accuracy of image registration in multi-sensor remote sensing images, changing mismatches into correct matches. Initially, key points matching is performed using the SIFT algorithm. Then, using the proposed affine-transformation-based approach, the mismatches are corrected and matching is done. Another novelty of the paper is suggesting two new criteria for assessing the efficiency of image matching methods in addition to the classical criteria of matching precision. As a weakness of the classical criteria that do not consider the total number of matches, feature repeatability rate and the number of correct matches are not defined efficiently. Simulation results show that the proposed method improves the rate of repeatability by 11.41% and cross- correlation coefficient by 14.20% on the average compared to the RANSAC method. Therefore, the proposed method can be used as a new and effective way of improving image matching.