Abstract
Face recognition is a personal identification system based on facial biometric data. By looking at the characteristics of the image data that has lighting variations and contains Gaussian/Poisson/Quantization noise when taken using a camera, facial recognition systems working in the real world with uncontrolled environmental conditions is a challenge. Local Ternary Pattern (LTP) is a representation of image features that are invariant to lighting, resistant to noise, but require manual threshold determination. In this study, we propose an automatic calculation of adaptive LTP’s threshold using the statistical characteristic of image histograms. We also propose a Multi-scale Block Adaptive Local Ternary Pattern (MBALTP) which combines local features extracted using adaptive LTP and global features captured using multi-scale block for face recognition. We experimented on the Extended Yale-B dataset, which were collected under different lighting variations and contains noise. The experimental result demonstrated that the proposed method can improve the recognition performance with an accuracy of 98.87%.
Original language | English |
---|---|
Pages (from-to) | 290-300 |
Number of pages | 11 |
Journal | International Journal of Intelligent Engineering and Systems |
Volume | 14 |
Issue number | 6 |
DOIs | |
Publication status | Published - Dec 2021 |
Keywords
- Adaptive threshold
- Face recognition
- Local ternary pattern
- Multiscale block