Face Recognition Based on a Single Image and Local Binary Pattern

This paper presents a novel algorithm for face recognition based on a single image and a new LBP (Local Binary Pattern) descriptor. The algorithm can be divided into three steps: firstly, calculating both the horizontal and vertical edge maps from the gray image; then extracting LBP histograms from those two edge images; finally, adopting elastic matching for classification. In addition, in order to reduce the LBP histogram's dimension which is too high, we propose a new LBP operator. It can reduce the dimension greatly and preserve the main information. The experimental results on standard face databases show that this algorithm performs better than traditional methods. It can obtain a higher recognition rate and is robust to posture, lighting and expressions.