Dimensionality reduced local directional pattern (DR-LDP) for face recognition

• Dimensionality reduced local directional pattern.
• Robust face recognition system.
• Highest recognition rate on benchmark datasets.

Local Directional Pattern (LDP) is a descriptor used for face recognition. It assigns a code for each pixel in the image, and the resultant LDP-encoded image is divided into regions for which each a histogram is generated. The histogram bins of all the regions are concatenated to form the final descriptor. In contrast to LDP, a dimensionality reduced local directional pattern (DR-LDP) is proposed in this paper. The proposed descriptor computes single code for each block by X-ORing the LDP codes obtained in a single block. During the process, restructuring of the patterns is done by slightly modifying the LDP coding pattern constraints. The significance of DR-LDP is the compact code generation for efficient face recognition. The experiments were carried out on standard databases like FERET, extended YALE-B database and ORL. The resultant DR-LDP descriptor provided better recognition rates, outperforming the existing local descriptor-based methods and proving its efficacy. The compact code can be further extended to provide biometric security.