一种Log-Gabor滤波结合特征融合的虹膜识别方法

提出了一种Log-Gabor滤波器结合特征融合的虹膜识别方法.该方法利用多尺度多方向二维Log-Gabor滤波器构成的多通道滤波器提取虹膜纹理特征,在特征层利用幅值信息对同尺度下多方向的Log-Gabor特征进行融合,以压缩冗余信息及去除非有效特征,然后对融合后的特征进行相位编码,并运用加权海明距离进行匹配,匹配时借助噪声屏蔽码去除眼睑遮挡干扰.提出了一种虹膜图像质量评价方法,可有效鉴别不适于识别的低质量虹膜图像.与传统的Gabor方法相比,新识别方法能得到更小的等错误率和相同错误接收率下更小的错误拒绝率,同时又将虹膜特征码的大小压缩为传统方法的1/2,可提高匹配速度及节约存储空间.在CASIA和UBIRIS虹膜库的测试结果表明,与传统Gabor方法相比,该方法在错误接受率为0.01%和0.1%时的错误拒绝率分别降低了0.57%和0.36%,等错误率降低了0.25%,特征码长度为128 B,减少了50%.

Iris Recognition Method Using Log-Gabor Filtering and Feature Fusion

An iris recognition method is developed based on Log-Gabor filtering and feature fusion.Multichannel 2D Log-Gabor filters are employed to extract the iris features.To reduce redundancy of the Log-Gabor features,multiple Log-Gabor features in the same scale with different orientations are combined by using the magnitude information,and the fusion features are encoded based on the phase information.The similarity of two iris codes is measured by their weighted Hamming distance.The noise mask codes are adopted to reduce the interference of eyelids occlusion.In addition,a method for iris image quality assessment is presented,which can discriminate the images that are unsuitable for recognition.The approach can achieve lower equal error rate and lower false rejection rate under same false acceptance rate,and the size of the iris codes is only a half of the traditional approach.The experimental results show that the false rejection rates at the false acceptance rates of 0.01%and 0.1% are respectively decreased by 0.57% and 0.36%,the equal error rate is decreased by 0.25%,and the size of iris codes is decreased by 50%,compared to the results of the traditional Gabor approach.