Abstract: A two-stage robust algorithm of singular points detection based on orientation features of fingerprint images is proposed. Candidate small blocks which may contain singular points are located with the integrated method of improved Poincaré index calculation with the orientation distribution analysis of the block orientation images; Local orientation variance in the candidate blocks is computed and used to locate the position of singular points accurately. The algorithm is evaluated with images from FVC2004 DB1_A database and the experimental results prove the method accurate and efficient.

Key words: Fingerprint identification, Singular point, Orientation image, Local orientation variance

摘要： 提出一种两阶段的奇异点检测方法。将指纹图像分块，求出各块的方向构成块方向图，并在块方向图的基础上利用邻域方向的分布分析结合改进的Poincare Index方法来确定奇异点所在的候选区域，对候选区域中的像素再通过计算局部方向变化率来确定奇异点的精确位置。将此方法用于对FVC2004 DB1_A指纹数据库的图像，实验结果表明这种方法对指纹图像中的噪声有很好的鲁棒性，并且计算简单快速，易于实现。

关键词: 指纹识别, 奇异点, 方向场, 局部方向变化率

CLC Number: 

• TP301