Limits...
Intensity Variation Normalization for Finger Vein Recognition Using Guided Filter Based Singe Scale Retinex.

Xie SJ, Lu Y, Yoon S, Yang J, Park DS - Sensors (Basel) (2015)

Bottom Line: However, the capacities and percentages of finger tissues (e.g., bone, muscle, ligament, water, fat, etc.) vary person by person.This usually causes poor quality of finger vein images, therefore degrading the performance of finger vein recognition systems (FVRSs).The experimental results on two public datasets demonstrate the effectiveness of the proposed method in enhancing the image quality and finger vein recognition accuracy.

View Article: PubMed Central - PubMed

Affiliation: Institute of Remote Sensing and Earth Science, College of Science, Hangzhou Normal University, Hangzhou 311121, China. shanj_x@hotmail.com.

ABSTRACT
Finger vein recognition has been considered one of the most promising biometrics for personal authentication. However, the capacities and percentages of finger tissues (e.g., bone, muscle, ligament, water, fat, etc.) vary person by person. This usually causes poor quality of finger vein images, therefore degrading the performance of finger vein recognition systems (FVRSs). In this paper, the intrinsic factors of finger tissue causing poor quality of finger vein images are analyzed, and an intensity variation (IV) normalization method using guided filter based single scale retinex (GFSSR) is proposed for finger vein image enhancement. The experimental results on two public datasets demonstrate the effectiveness of the proposed method in enhancing the image quality and finger vein recognition accuracy.

Show MeSH
ROC curves of different image enhancement methods on MMCBNU_6000 using different feature extraction algorithms: (a) DWT method; (b) LBP method; and (c) LPQ method.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4541924&req=5

sensors-15-17089-f011: ROC curves of different image enhancement methods on MMCBNU_6000 using different feature extraction algorithms: (a) DWT method; (b) LBP method; and (c) LPQ method.

Mentions: The receiver operating characteristic (ROC) curves of the different image enhancement methods with different algorithms for feature extraction are depicted in Figure 10. The EER values and the improved ratios calculated based on the EER values obtained using original image are listed in Table 1. The positive ratios represent the corresponding image enhancement algorithm can enhance the matching accuracy while the negative ratios denote the image enhancement method can degrade the matching accuracy. It is clearly shown that SSR, which is effective for facial image illumination normalization, is not beneficial for enhancing the quality of finger vein when using DWT and LPQ for feature extraction. HE exacerbates the poor image contrast in some local regions, as shown in Figure 10e, and thus cannot enhance the matching accuracy. In contrast, SR can enhance the performance a little by using DWT and LPQ for feature extraction. Note that different feature extraction algorithms focus on extracting different kinds of features from an image and different image enhancement methods aims at enhancing images in various styles. Thus, as listed in Table 1, DWT, LBP, and LPQ gave various matching accuracies on different enhanced images. SSR is beneficial for image enhancement when using LBP for feature extraction, but it fails with usage of DWT and LPQ. A similar circumstance appears for images enhanced using IN. These cases demonstrate that IN, SR, SSR and HE are not stable for enhancing image quality. However, it is shown in Figure 11 and Table 1 that the proposed GFSSR has the best performance, no matter which feature extraction was adopted. It results in making lowest EER values, which reduces EER values by 34.6%, 31.8%, and 33.0% compared to the EER values obtained on the original images, using DWT, LBP, and LPQ, respectively.


Intensity Variation Normalization for Finger Vein Recognition Using Guided Filter Based Singe Scale Retinex.

Xie SJ, Lu Y, Yoon S, Yang J, Park DS - Sensors (Basel) (2015)

ROC curves of different image enhancement methods on MMCBNU_6000 using different feature extraction algorithms: (a) DWT method; (b) LBP method; and (c) LPQ method.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4541924&req=5

sensors-15-17089-f011: ROC curves of different image enhancement methods on MMCBNU_6000 using different feature extraction algorithms: (a) DWT method; (b) LBP method; and (c) LPQ method.
Mentions: The receiver operating characteristic (ROC) curves of the different image enhancement methods with different algorithms for feature extraction are depicted in Figure 10. The EER values and the improved ratios calculated based on the EER values obtained using original image are listed in Table 1. The positive ratios represent the corresponding image enhancement algorithm can enhance the matching accuracy while the negative ratios denote the image enhancement method can degrade the matching accuracy. It is clearly shown that SSR, which is effective for facial image illumination normalization, is not beneficial for enhancing the quality of finger vein when using DWT and LPQ for feature extraction. HE exacerbates the poor image contrast in some local regions, as shown in Figure 10e, and thus cannot enhance the matching accuracy. In contrast, SR can enhance the performance a little by using DWT and LPQ for feature extraction. Note that different feature extraction algorithms focus on extracting different kinds of features from an image and different image enhancement methods aims at enhancing images in various styles. Thus, as listed in Table 1, DWT, LBP, and LPQ gave various matching accuracies on different enhanced images. SSR is beneficial for image enhancement when using LBP for feature extraction, but it fails with usage of DWT and LPQ. A similar circumstance appears for images enhanced using IN. These cases demonstrate that IN, SR, SSR and HE are not stable for enhancing image quality. However, it is shown in Figure 11 and Table 1 that the proposed GFSSR has the best performance, no matter which feature extraction was adopted. It results in making lowest EER values, which reduces EER values by 34.6%, 31.8%, and 33.0% compared to the EER values obtained on the original images, using DWT, LBP, and LPQ, respectively.

Bottom Line: However, the capacities and percentages of finger tissues (e.g., bone, muscle, ligament, water, fat, etc.) vary person by person.This usually causes poor quality of finger vein images, therefore degrading the performance of finger vein recognition systems (FVRSs).The experimental results on two public datasets demonstrate the effectiveness of the proposed method in enhancing the image quality and finger vein recognition accuracy.

View Article: PubMed Central - PubMed

Affiliation: Institute of Remote Sensing and Earth Science, College of Science, Hangzhou Normal University, Hangzhou 311121, China. shanj_x@hotmail.com.

ABSTRACT
Finger vein recognition has been considered one of the most promising biometrics for personal authentication. However, the capacities and percentages of finger tissues (e.g., bone, muscle, ligament, water, fat, etc.) vary person by person. This usually causes poor quality of finger vein images, therefore degrading the performance of finger vein recognition systems (FVRSs). In this paper, the intrinsic factors of finger tissue causing poor quality of finger vein images are analyzed, and an intensity variation (IV) normalization method using guided filter based single scale retinex (GFSSR) is proposed for finger vein image enhancement. The experimental results on two public datasets demonstrate the effectiveness of the proposed method in enhancing the image quality and finger vein recognition accuracy.

Show MeSH