A hierarchical identification method based on improved hand geometry and regional content feature for low-resolution hand images

Abstract

Low-quality images are taking an increasing amount in real-life biometrics identification systems, especially when the working environment is gradually becoming complex and open. Low resolution is a crucial component of low quality. We propose a hierarchical identification method based on improved hand geometry and regional content feature for low-resolution hand images without region of interest (ROI) cropping. At coarse level, angle information is added as a complement to line-based hand geometry. At fine level, relying on the assumption that gradient value of each pixel presents the gray-level changing rate, we develop a simple sequence labeling segmentation method, and choose conditional regions that are relatively steady in segmentation through region area constraint. Because distinctive lines and dense textures always have lower gray-levels than their surrounding areas, regions with lower average gray-levels are selected from conditional regions. Regional centroid coordinates are extracted as feature vectors. Finally, regional spatiality relationship matrix is built up to measure distances between feature vectors with various dimensions. Compared with ROI-based state-of-the-art algorithms, the proposed method utilizes more available information in low-resolution hand images. Performance comparisons between algorithms prove that our method is effective and robust to rotation and translation.