|
Improving iris verification systems through partial iris template comparison |
|---|---|
| รหัสดีโอไอ | |
| Title | Improving iris verification systems through partial iris template comparison |
| Creator | Intouch Wangtrakoondee |
| Contributor | Waree Kongprawechnon, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2568 |
| Keyword | Iris authentication, Iris pattern recognition, Partial iris template, Robust iris template, Template matching |
| Abstract | Iris is regarded as one of the most secure physical identifiers in verification systems. An important step in iris verification involves isolating the iris from the periocular region before normalizing it into a rectangular image. However, this normalization process is highly susceptible to distortions caused by variations in detected iris at various angles. To address this issue, this thesis proposes a quantized masking technique designed to filter out distorted, and subsequently unreliable regions of the iris template. The methodology was evaluated across multiple datasets utilizing two distinct segmentation approaches: SAM segmentation proposed by Chokchaisiri et al. (2024) and the Two-headed segmentation method proposed by Lazarski et al. (2022). The experimental results demonstrate consistent performance enhancements across all datasets and segmentation methods, notably reducing distribution overlap and the precision of the system. |
