基于三方向图的多尺度平滑指纹奇异点检测

Multi-scale smooth and fingerprint singularity detection based on three-directional image

为解决低质量指纹图像定位奇异点位置偏移及定位区域过大的问题,提出了一种基于三方向图的多尺度平滑奇异点检测算法.首先,计算指纹图像的方向场和频率场估计并用Gabor滤波器对指纹图像进行增强.然后,计算指纹图像中脊线上像素点的方向并对其进行三方向划分得到指纹脊线方向图.最终,去掉噪点并填补非脊线区域,同时检测多个平滑尺度下的奇异点位置并通过信息融合剔除伪特征点,精确定位奇异点.对FVC2002,FVC2004和FVC2006指纹图像库进行实验,结果表明:该算法对低质量指纹图像奇异点检测的准确率有明显提升,奇异点检测的总准确率达92.35%,可将奇异点精确定位在2×2像素区域内.

In order to solve the issues of singularity position offset or inaccurate singularity detection in low quality fingerprint images, a method on multi-scale smooth and fingerprint singularity detection based on the three-directional image was proposed.Firstly,the directional field and the frequency estimation were calculated and a Gabor filter was applied to enhance the low quality fingerprint image.Subsequently,the pixel direction of fingerprint ridge was computed,then the directional image of fingerprint ridge was obtained by dividing the directions into three parts.Finally,after eliminating the noise and completing the non-ridge area,singularities under multi-scales of smoothing were detected.Furthermore,information fusion was applied to eliminate the false feature points to accurately locate the singularities.Experimental results on FVC2002,FVC2004 and FVC2006 show that the accuracy of singularity detection obtained by the proposed method is improved and the accuracy ratio is 92.35%.Furthermore,the proposed method can precisely locate the singularities in a 2×2 pixels area.