Abstract: To improve the blur edge of micro electromechanical systems(MEMS) microstructures image caused by filtering the noise, an improved anisotropic SUSAN filtering algorithm was proposed. In this method, the variances of long axis and short axis were determined in accordance with the independent density spread model, the long axis direction was determined by the gradient direction of the point and the SUSAN filter threshold was determined by the partial average difference between gray value of local image and nucleus. So anisotropic SUSAN filter was constructed. The filter can maintain the edge features as well as process the image smoothing. The experimental results show that anisotropic SUSAN filter could reduce the noise and preserve the edge information at the same time.