关键词: 脉冲耦合神经网络, 指纹图像, 细化算法, 三角模板, 方形模板

Abstract: There is much trouble in thinning algorithm based on traditional PCNN，such as not thinning to one pixel，breaking the connectivity of ridge，generating spikes and so on.In order to overcome these disadvantages，an improved binary fingerprint image thinning algorithm based on parallel capability of PCNN is proposed.By analyzing character of spikes，some templates are proposed.Using these templates，the spikes and breaks can effectively be eliminated.Also，this method can effectively solve the less thorough thinning problem.The algorithm tests on FVC2004 database.The results show that the method can effectively reduce breaks and spikes，and apply to the automatic fingerprint recognition system to improve accept rate.

Key words: Pulse Coupled Neural Networks（PCNN）, fingerprint image, thinning algorithm, triangle templates, rectangle templates