A robust feature matching algorithm based on CSIFT descriptors

Feature matching is an essential and crucial step in image registration procedure. Since accurate and efficient image registration approach is still a challenging problem in image processing, with the aim of handling this kind of problem, this paper presents a robust feature matching algorithm based on CSIFT descriptors. Comparing with the traditional feature descriptors, CSIFT (colored invariant feature transform) descriptors can make use of the color information of images, which prove to be more efficient and robust in feature matching. After describing CSIFT descriptors, we proposed an improved feature matching algorithm. Inlier features with the highest frequency are selected as the input to calculate the matching Homography Matrix. Experimental results indicate that our method is effective and robust.