Efficient face recognition based on GAP feature

In this paper, we propose efficient face recognition based on Grayscale Arranging Pairs (GAP) feature. GAP is a robust holistic feature considering the intensity relationships about all of pixels. Therefore, it has good performance for face recognition. However, GAP feature consider all of pixels and it takes high computational time. In order to reduce computational time, we uses GAP feature in terms of dominant parts such as eyes, nose, mouth and so on. Considering dominant pixels reduces computational time as well as maintains recognition performance. In experiment result, we compare the proposed method with GAP feature in Extended Yale B database. This dataset has various illuminations but the recognition performance of the proposed method has good performance with low computational time.