کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Texture description and Face Recognition using Weighted Local Patch Distance Vectors

نویسندگان :

Ziba Javanmardi¹ Farzam Mohebbi² Seyed Saeed Hayati³

1- دانشگاه علوم و فنون دریایی خرمشهر 2- دانشگاه علوم و فنون دریایی خرمشهر 3- دانشگاه علوم و فنون دریایی خرمشهر

کلمات کلیدی :

texture description،face recognition،image feature extraction،scale invariance،robustness to noise

چکیده :

Texture-based descriptors are integral to advancing face recognition systems, especially under varying conditions of illumination, noise, and scale. This paper proposes an enhancement to the Local Patch Distance Vector (LPDV) texture descriptor, termed Weighted Local Patch Distance Vector (WLPDV). By incorporating weighted Euclidean distances derived from patch variances, the proposed method improves the sensitivity of LPDV to critical texture details. Furthermore, the integration of a scalespace framework enhances scale invariance. Extensive evaluations on texture and face datasets, including Outex, Brodatz and CelebA, demonstrate the robustness of WLPDV against noise, rotation, illumination, and scale variations. Results reveal superior performance compared to existing descriptors such as LBP, LTP, and CLBP, establishing WLPDV as a versatile and efficient tool for both texture description and face recognition tasks.