Improving the accuracy of verification using fuzzy minutiae-based matching

Fingerprint images usually have the rare perfect quality. They may be degraded and corrupted with noises because many factors, such as skin and impression conditions, vary. Many methods and algorithms (e.g., preprocessing, postprocessing, and matching methods) are studied for a long time. Among those matching methods, the most well-known matching methods are correlation-based, ridge-feature based, minutiae-based method. This paper only discusses minutiae-based matching method. The previous minutiae-based matching method has the tolerance values that compensate for these unavoidable errors and distortions. Global adaptive threshold that we use for the minutiae-based matching is not efficient to improve the accuracy because the values less than the threshold have the same results. Thus, we use fuzzy membership function to represent the difference between minutiae. In this paper, we present the fuzzy minutiae-based matching method to improve the accuracy of the verification between template and input fingerprint images. All the methods and algorithms described in this paper are implemented using Visual C++ 6.0 and we can reduce the false reject rate by approximately 7% at a false acceptance rate of 1%.