Abstract: This paper proposes a novel fingerprint recognition algorithm based on Gabor filterbank to extract both global and local features of the fingerprints.At present，the widely used minutiae-based algorithm performs well on recognition while the preprocessing is time-consuming.On the other hand，the traditional structure-based algorithm is highly efficient but performs poorly on rotated pictures.The proposed approach uses filters to locate the core point，then tessellates the feature region centered at the reference point with 16-direction-sectors.Finally the features of both original picture and the rotated picture are extracted respectively through the 8-direction Gabor filtering.The experimental results on the FVC2004 fingerprint database show the good performance of the method on both accuracy and efficiency.

摘要： 提出了一种基于Gabor滤波提取指纹全局及局部特征的识别算法。目前广泛使用的基于细节点的算法识别性能较高，但由于需要前期预处理，增加了系统开销。另一方面，传统的基于结构特征的算法速度较快，然而对偏转指纹的识别性能较差。针对以上不足，利用滤波器定位参考点，并将以参考点为中心的特征提取区域划分为16方向的扇区，然后通过8方向滤波分别提取原始图像和旋转后图像的特征。在FVC2004指纹库上的对比实验证明，该算法同时具备了较好的运行效率和识别性能。