提出一整套手指静脉认证系统的图像处理和静脉模板匹配算法。算法采用Sobel 算子结合首遇跟踪法检测手指边缘，然后采用最小均方误差原则拟合手指边缘并对图像进行旋转校正归一化；基于最大曲率模型的思想，采用四个方向梯度算子提取静脉；为降低拒真率和提高匹配速度，对静脉图像进行膨胀、三值化和两次压缩，建立包含静脉、背景和模糊区的紧凑型指静脉模板，并将测试模板与登记模板进行交叉点对点的匹配。实验结果表明，该算法兼顾了效率和准确性，只要手指没有反转，就可以得到接近100%的识别率，与传统匹配算法相比在速度和有效性方面都具有明显优势。

A full set of algorithms on image processing and recognition for finger vein authentication was proposed. A new image normalization algorithm was designed. During normalization, a Sobel operator incorporating the first meet tracing was used to detect finger edges, and then image correction with rotation was done after the finger edges were fitted under the least mean-square error principle. Based on maximum curvature models, finger vein was extracted by using four direction gradient operators. To decrease False Rejection Rate (FRR) and increase recognition rapid, after expanded, the image was segmented with three levels and compressed to become a more compact vein pattern including finger vein, background and vague parts. At last, a new matching algorithm with point to point between mask and test patterns was put forward. Our experiments verify that our methods outperform traditional ones on speed and effectiveness. Its recognition rate is up to 100%, with a finger set horizontally.