Abstract: This paper presents a global reference point detection algorithm of fingerprint images. After the downward-curvature intensity graph is obtained, the algorithm can locate the reference point from where a maximum line-projection starts. Another algorithm for estimating of supposed reference point which is outside the foreground has been also proposed. Experiment results show that this algorithm has a higher accuracy than that of traditional Poincare Index algorithm, and it is also less sensitive to noise than traditional Poincare Index algorithm. Moreover, this algorithm can deal with plain-arch type fingerprints which Poincare Index can’t.

Key words: reference point, singular points, fingerprint recognition, orientation field

摘要： 提出了一种全局指纹图像参考点检测算法，首先计算指纹的下曲率值，然后建立一个参考点检测模型，利用下曲率最大的优化方法确定参考点。还提出了一种估计参考点的算法，当参考点在图像的外面时，算法可以估计出参考点的位置。实验结果显示，算法不仅精度比传统的PI算法高，而且也比PI稳健，可以检测出拱形指纹的参考点，而PI算法则不能。

关键词: 参考点, 奇异点, 指纹识别, 方向场