Automatic fingerprint identification is one of the most important biometric technologies. Automatic fingerprint identification depends mainly on the comparison of ridge characteristics. The ridge characteristics include ridge ending, ridge bifurcation. How to extract those minutiae effectively is the key of automatic fingerprint identification. This paper uses a square Gabor filter using local average frequency. The method can overcome the blocky effect of Lin Hong's Gabor filter. This paper proposes a new evaluation GI(Goodness Index) method of the fingerprint image enhancement algorithm. This evaluation GI method has a wide prospect.