Novel remote sensing image registration method based on the improved SIFT descriptor

The quantity and quality of features detected from images are important to feature-based registration. The SIFT descriptor has been proven to be one of the most robust local invariant feature to common geometric deformations. To enhance the effectiveness of this method, this paper focuses on studying an improvement version of the SIFT descriptor via extending the gray level distribution automatically at the cost of discarding the points with levels approach the maximum and minimum of the whole grey level, and applying it to image registration with the robust estimator method of PERANSAC and a stricter NND threshold. Experimental results demonstrated that the proposed approach can not only add the descriptors extracted with the same parameters, but also improved the match ratio simultaneously. The improved method is also robust to different image transformations and matching threshold.