Deformed Palmprint Matching Based on Stable Regions.

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Palmprint recognition (PR) is a reliable biometric technology.
• Image deformations significantly degrade PR performance, especially in contactless scenarios.
• Existing methods struggle with non-linear palmprint deformations.

Purpose of the Study:

• To develop a robust palmprint matching model for non-linearly deformed images.
• To propose a novel approach, Key Point-based Block Growing (KPBG), to address deformation challenges in PR.
• To enhance the applicability of PR in unconstrained environments.

Main Methods:

• A model approximating non-linear deformations using piecewise-linear stable regions was derived.
• The Key Point-based Block Growing (KPBG) algorithm was developed for deformed palmprint matching.
• Scale Invariant Feature Transform (SIFT) features and an iterative M-estimator sample consensus algorithm were used to approximate transformations.

Main Results:

• The proposed model effectively handles non-linear palmprint deformations.
• KPBG demonstrated superior performance in palmprint verification compared to state-of-the-art methods.
• Experiments on public databases validated the effectiveness of the KPBG approach.

Conclusions:

• The developed model and KPBG algorithm offer a robust solution for deformed palmprint matching.
• This research significantly improves the accuracy and reliability of contactless palmprint recognition.
• The findings pave the way for wider applications of PR in various security systems.