激光与光电子学进展, 2021, 58 (2): 0210019, 网络出版: 2021-01-11
基于NSST和Tetrolet能量特征的指关节纹识别 下载: 719次
Finger-Knuckle-Print Recognition Based on NSST and Tetrolet Energy Features
图像处理 指关节纹识别 能量曲面 非下采样的Shearlet变换 Tetrolet变换 image processing finger-knuckle-print recognition energy surface non-subsampled Shearlet transform Tetrolet transform
摘要
针对目前指关节纹识别方法鲁棒性差的问题,提出了一种基于非下采样的Shearlet变换(NSST)和Tetrolet能量特征的指关节纹识别方法。首先,采用直方图均衡化调整图像的灰度,以减少光照分布不均对识别系统产生的影响。其次,利用NSST及其逆变换得到去噪后的重构图像,并对其进行Tetrolet变换,建立低频图像的能量曲面。最后,将不同图像的能量曲面作差,得到能量差曲面,进一步计算曲面的方差,并以此为依据对不同指关节纹图像进行分类识别。在HKPU-FKP、IIT Delhi-FK、和HKPU-CFK图库及其噪声图库的实验结果表明,本方法的正确识别率可达98.0392%,最短识别时间为0.0497 s,最低等误率为2.5646%。相比其他方法,本方法可以明显提高指关节纹识别系统的性能,具有可行性和有效性。
Abstract
Aiming at the identification problem with poor robustness based on current finger-knuckle-print recognition methods, a finger-knuckle-print recognition method using non-subsampled Shearlet transform (NSST) and Tetrolet energy features is proposed in this paper. First, histogram equalization is used to adjust the gray level of the image to reduce the influence of uneven light distribution on the recognition system. Second, the NSST and its inverse transform are used to obtain the reconstructed image after denoising, and Tetrolet transform is performed on it to establish the energy surface of low-frequency image. Finally, the energy difference surface is obtained by subtracting the energy surface of different images, and the variance of the surface is further calculated. Based on this, the classification and recognition of different finger joint print images are carried out. The experiment results on the HKPU-FKP, IIT Delhi-FK, and HKPU-CFK databases and their noise databases show that the correct recognition rate of the method is 98.0392%, the shortest recognition time is 0.0497 s, and the lowest equal error rate is 2.5646%. Compared with other algorithms, the algorithm improves the performance of the finger-knuckle-print recognition system, which is feasible and effective.
王媛, 林森. 基于NSST和Tetrolet能量特征的指关节纹识别[J]. 激光与光电子学进展, 2021, 58(2): 0210019. Yuan Wang, Sen Lin. Finger-Knuckle-Print Recognition Based on NSST and Tetrolet Energy Features[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0210019.