Abstract: To resolve the problem of low correct matching rate in IR/visual image matching,an improved approach to reject mismatches based on support vector regression(SVR) is proposed.Supposing that the pairs of control points were known,the method applied SVR to estimate the point correspondence function;then based on the function to judge whether matches are right or not.Experiment results indicate that the proposed method is a feasible approach for rejecting mismatches.Compared with RANSAC,it can improve correct matching rate under the condition of remaining the original right matching.