Abstract: In order to solve the problem of self-adaptive and accurate segmentation of knuckleprint，this paper brings forward a new segmentation algorithm of knuckleprint based on modified Gabor filter.In the stage of calculation of character direction，this algorithm uses the direction of finger axis as Gbor filter’s direction，which solves the problem of rotation overflow and singular value region in traditional Gabor filter.In the segmentation stage，this algorithm uses certain frequence value to processing characteristics of knuckleprint，which can reduce time-consuming and avoid to calculate with certain projection value of Gbor parameters.Emulation experiment shows that this algorithm not only reduce time-consuming，get rid of noise and hand-shape affection but also robust in different quality images.

Key words: knuckleprint, Gabor filter, finger axis, direction images, frequence projection

摘要： 针对指横纹难于实现自适应分割的问题，提出一种新的基于改进Gabor滤波器的指横纹提取算法。该方法在指横纹特征方向求取阶段，为避免出现传统Gabor滤波器固有的旋转溢出和奇异值区间的问题，提出以手指轴线方向作为Gabor滤波器方向。在分割阶段，为减少求取Gabor频率的耗时和Gabor频率投影值难于确定的问题，提出以固定频率对指横纹特征进行Gabor滤波。实验证明该算法不仅能够减少计算量，还能有效克服噪声以及手形姿态变化的影响，并且对不同精度的指横纹图像都有较强的鲁棒性。

关键词: 指横纹, Gabor滤波器, 手指轴线, 方向图, 频率投影