基于LBP和Gabor方向的韦伯局部描述子的指静脉识别

Finger Vein Recognition Based on LBP and Gabor Direction Weber Local Descriptors

韦伯局部描述子(WLD)的差分激励存在正负抵消现象且对噪声敏感,难以准确地提取指静脉的纹理特征,所采用的梯度方向算子仅包括水平和垂直方向,无法有效地反映指静脉的方向特征.根据指静脉图像具有丰富纹理信息和方向特性的特点,对WLD中的差分激励和梯度方向分别进行改进.首先,在计算局部区域的灰度差时,采用均匀模式LBP替代原始差分激励,有效地提取了指静脉图像的局部纹理特征,同时减少噪声的干扰;然后,采用Gabor方向代替原始梯度方向,有效地刻画指静脉曲线的方向特征.在FV-TJ和FV-USM指静脉库上的实验结果表明:与其他局部描述子算法相比,所提出方法有更高的识别率和更低的错误率.

The differential excitation of Weber’s local descriptor(WLD)has positive and negative cancellation and is sensitive to noise,so it is difficult to extract the texture features of the finger veins accurately,and the gradient direction operator used only includes the horizontal and vertical directions,which cannot effectively reflect the direction of the finger veins feature.Therefore,according to the characteristics of rich vein information and directional characteristics of finger vein images,the differential excitation and gradient directions in WLD are improved respectively.First,when calculating the gray level difference in a local area,uniform mode LBP is used instead of the original differential excitation,which effectively extracts the local texture features of the finger vein image while reducing noise interference;second,Gabor direction is used to replace the original gradient direction to effectively describe the direction characteristics of finger vein curve.Experimental results on FV-TJ and FV-USM finger vein libraries show that the proposed method has higher recognition rate and lower equal error rate than other local descriptor algorithms.