A Method for Face Recognition Based on Rotational Invariant Structural Texture Primitives (RISTP) Derived on Local Directional Pattern (LDP)

摘要

This paper derives a new way of extracting Bezier and Koch curves and other topological structures on local directional pattern (LDP) facial images. The curve and topological structures on facial image textures will appear more significantly and provides an efficient framework for analyzing object shape characteristics (such as size and connectivity) due to its geometry-oriented nature. This paper transforms the raw facial image into a more stable code by deriving LDP code, which is consistent in the presence of noise and illumination changes, since edge response magnitude is more stable than pixel intensity. This paper divided the LDP coded facial image in to stable edge response regions of size 5×5 and 3×3 and derived curve and other structural features respectively for efficient face recognition. The proposed geometrical/structural aspects on LDP are rotational and pose invariant when compared to pattern trends that represent a shape. The proposed method is compared with the other state of art local based approaches on popular databases. The results indicate the efficacy of the proposed method.