Improved Local Line Binary Pattern (Illbp): An Improved Lbp-Based Biometric Descriptor For Face And Finger Vein Recognition
Face recognition under different illumination remains a challenging problem. The variations between the images of the same face due to illuminations are almost always being larger than image variations due to changes in face identity. For finger vein recognition, the recognition rate may be de...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2013
|
| 主題: | |
| 在線閱讀: | http://eprints.usm.my/45063/1/Chaiwuh%20Shing24.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Face recognition under different illumination remains a challenging problem. The
variations between the images of the same face due to illuminations are almost always
being larger than image variations due to changes in face identity. For finger vein recognition,
the recognition rate may be degraded due to low quality of finger vein images.
This is because finger vein images are not always clear and can display irregular shadings.
A theoretically simple, yet efficient technique, called Improved Local Line Binary
Pattern (ILLBP) has been proposed in order to solve the problems. The descriptor can be
used for both face and finger vein recognition. The effectiveness of the proposed technique
is empirically demonstrated using Principal Component Analysis-k-Nearest Neighbor
(PCA-kNN), Multiclass Support Vector Machine (Multiclass SVM) and Hamming
Distance(HD) as the classifiers. Comparisons among other existing Local Binary Pattern
(LBP) variants on the Yale Face Database B, Extended Yale Face Database B and our own
finger vein database have been conducted. The advantages of our technique include higher
accuracy compared to other LBP variants and fast computational time. The experimental
results for face recognition showed that by using PCA-kNN, the best ILLBP (N = 15, P
= 2) achieved a high recognition rate (89.24%) only slightly worse than the best LLBP
with N = 17 (89.36%). |
|---|