Face detection by shape context (Thai)

Abstract

At present, human identification using face recognition is a challenge topic. The efficiency of face
detection will have influence on the performance result of face recognition. This research presents the face
detection from varying background images. In this paper, we applied skin detection combined with a
shape context in order to improve the performance of face detection. We proposed technique includes 1) to detect skin colors from HSV and YCbCr standard color models to get a skin region, 2) the skin region in step 1, we detect of region 3) to confirm the detected face through assessing between the edge of region and face template using a shape context. In this research, we experimented on 82 color images. The experimental results show that the research has high performance, with an average accuracy rate of 0.85.