A new local pattern method using directions and subpattern counting for face classification.
Jeerayut Wetweerapong, Supatchariya Sichana, Pikul Puphasuk · International journal of mathematics and computer science · 2025
Image classification plays a crucial role in modern intelligent vision applications by extracting both local and global features from images to identify and categorize them into predefined classes. In this paper, we propose a new local pattern method that counts the subpatterns of three comparison values along four directions (LP4D) for face classification. It extends the traditional Local Binary Pattern (LBP) method by incorporating directional information and subpattern counting to construct feature vectors and their corresponding histograms. We evaluate the performance of LP4D against LBP and its improved variants using several face databases. Experimental results demonstrate that LP4D overall outperforms the compared methods.