基于稀疏结构噪声检测的指静脉图像去噪算法

针对携带污染噪声的指静脉图像中背景区域、静脉区域和噪声区域的稀疏特性，提出一种改进的指静脉图像去噪算法。利用指静脉稀疏结构特性建立鲁棒主成分分析（RPCA）模型，通过交替方向乘子法求解RPCA模型获得含稀疏目标的前景图像并对其进行阈值分割以提取噪声分布图，同时根据提取结果建立修复优先度规则和自适应选择性滤波模板，实现指静脉图像的去噪处理。实验结果表明，与自适应非局部均值去噪算法和基于分数阶微分梯度噪声检测的去噪算法相比，在零误识情况下该算法处理后的带噪指静脉图像拒识率平均降低5.95%和3.64%，有效提升了带噪指静脉图像的识别性能。

Denoising Algorithm for Finger Vein Images Based on Sparse Structure Noise Detection

Aiming at the sparse characteristics of the background area,vein area and noise area in the finger vein image with pollution noise,this paper proposes an improved denoising algorithm for finger vein images.Based on the sparse structure of finger veins,a Robust Principal Component Analysis(RPCA)model is established,and solved by using the Alternating Direction Method of Multipliers(ADMM)to get the foreground image with sparse targets.For the obtained image,threshold segmentation is implemented to extract the noise distribution image,and the repair priority rule and adaptive selective filter template based on the extraction result are established to remove noise.Experimental results show that in the case of zero FAR,the FRR of the finger vein images processed by this algorithm is reduced by 5.95%and 3.64%respectively compared with processed by the Adaptive Non-Local Means(ANLM)denoising algorithm,and processed by denoising algorithm based on fractional order Fractional Gradient(FG)noise detection.The proposed algorithm effectively improves the recognition performance of noisy finger vein images.