Descriptor feature based on local binary pattern for face classification

Abstract

Local Binary Patterns (LBP) is a non-parametric descriptor whose purpose is to effectively summarize local image configurations. It has generated increasing interest in many aspects including facial image analysis, vision detection, facial expression analysis, demographic classification, etc. in recent years and has proven useful in various applications. This paper presents a local binary pattern based face recognition (LBP) technology using a Vector Support Machine (SVM). Combine the local characteristics of LBP with universal characteristics so that the general picture characteristics are more robust. To reduce dimension and maximize discrimination, super vector machines (SVM) are used. Screened and Evaluated (FAR), FARR and Accuracy Score (Acc), not only on the Yale Face database but also on the expanded Yale Face Database B datasets, the test results indicate that the approach is accurate and practical, and gives a recognition rate of 98 %.

Ortho-LBP: Novel Features for Enhanced Pattern Recognition through Bitwise Logical Fusion of Bidirectional Orthogonal Information with LBP

“…Compared to traditional methods like LBPriu2, LBPOUS demonstrates superior performance on texture datasets like Outex and CUReT, highlighting the importance of innovative LBP variants in advancing texture analysis techniques. Sahy et al [19] presented a method for robust face recognition by integrating LBP with a Vector Support Machine (SVM). Through experiments on the Yale Face database, the method achieves impressive recognition rates, showcasing the efficacy of combining local and global features for accurate recognition across diverse datasets.…”