This paper proposes a registration method for microscopic displacement image based on Speeded Up Robust Features(SURF) algorithm. This method is designed to solve problems of high mismatch ratio in microscopic displacement and large time computation caused by extracting too many feature points. Taking the characteristics of the microscopic displacement experimental model into account, this method extracts the boundary of the image containing rich amount of information, and extracts feature points at the boundary of image. The feature points which meet special conditions are obtained by revised Hassion corner detection method. Experimental results illustrate that the proposed method can improve correct matching rate and reduce time computation at the same time.