According to the motion characters of MEMS micro-structure,an optical algorithm based on label field and neighborhood optimization is presented. The method is more suitable than classical Horn-Schunk optical algorithm to motion estimation for micro-structure. A set of optical testing system is developed to visualize the motions of micro-structures. This system included a light micro-scope to magnify the image of the micro-structure and project the image ot the CCD camera; froze the high speed motion station of micro-structure based on stroboscopic principle to get a serial of dynamic images;and used the optical algorithm based on label field and neighborhood optimization method to motion estimation for sequential images of micro-structures to obtain dynamic characteristics parameters. The proposed algorithm has distinguishing features of nonconnect and high measurement precision. Experimental results indicate that the mean standard deviation of measurement repetition is 40 nm