Summary: Thinning is a very important process in Automatic Fingerprint Identification Systems (AFIS). But the existing thinning algorithms have many problems such as thinning not to one pixel, ridge swallowing, thinning skeleton not in ridge center and so on. Through studying and analyzing the algorithm of image post processing, an improved image template thinning algorithm is proposed based on a group of modified templates. The experimental result prove that the algorithm is completed, and the refined fingerprint skeleton is in the middle line of grain line and burr-free. Compared with the traditional refinement algorithm, the refinement result is closer to the fingerprint grain line center, so it is convenient for characteristic extracting and it reduces the matching complexity.