用于非接触式指纹增强的曲率驱动自适应滤波器  

Adaptive curvature-driven filter for contactless fingerprint enhancement

作　　者：李海燕[1] 王唐宇 郭磊[1] 余鹏飞[1] Li Haiyan Wang Tangyu Guo Lei Yu Pengfei(School of Information Science, Yunnan University, Kunming 650040, Chin)

机构地区：[1]云南大学信息学院

出　　处：《华中科技大学学报（自然科学版）》2018年第2期128-132,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61561050,61462094)

摘　　要：为了有效增强指纹,提出了一种改进的基于曲率的自适应非接触式指纹图像增强算法.首先,对局部平稳的指纹图像信号进行STFT（short time Fourier transform）分析,确定纹理模式频谱所在的主要区域,对频谱进行根滤波和形态学平滑后,用均值-门限法提取STFT主频谱成分;然后,提出一种角度和径向估计方法,计算指纹图像在频域的方向图和频率图;最后,根据指纹的曲率,构造相应的巴特沃斯陷波带通滤波器和高斯带通滤波器分别对低曲率区域和高曲率区域进行滤波,有效地滤除奇异点区域和非奇异点区域的噪声模式频率并保留纹线模式的主频率.将提出算法和同类算法进行主客观的实验比较,实验结果表明：提出的算法优于现有算法,能有效滤除非接触式指纹的噪声、增强脊线和谷线之间对比度,同时修复并增强正确的脊线结构.In order to effectively enhance fingerprint image,an adaptive fingerprint enhance algorithm based on the curvature was proposed.Firstly,the fingerprint image was considered as a non-stationary 2 D signal and was transformed to Fourier domain by short time Fourier transform analysis. Subsequently, a creative angular and radial estimation method was proposed to calculate the orientation and the frequency images of the fingerprint in Fourier domain after performing root filtering and morphological filtering.Finally,based on the curvature of fingerprint,a Butterworth notch band pass filter and a Gaussian band pass filter were applied to filter the low curvature area and the high curvature area, respectively, which can effectively remove the noise in the singular area and non-singular area while remaining the main frequency pattern. The proposed method was subjectively and objectively compared with approaches in literature.The results demonstrate that the proposed method is of better performance than the compared methods since it can remove noise,enhance the contrast between ridge and valley of fingerprint and simultaneously preserve the correct ridge direction.

关 键 词：非接触式指纹图像 STFT分析 巴特沃斯陷波带通滤波器 高斯带通滤波器 图像增强算法 

分 类 号：TN929.5[电子电信—通信与信息系统]