Robust extraction of face candidate through segmentation and conditional merging in skin area

Abstract

In the face detection, using skin color information can reduce the search range for a face candidate. However, in a complex scene, the face detection based on using skin color failed when it deals with skin color area including similar color background, overlapping faces and factual face divided into multi areas. The aim of this work is focus on robust extraction of face candidate for face detection system. Firstly, the areas of skin in the input image are detected by the skin color classifier, holes filter, and morphological operations. Secondly, it uses image segmentation technique to divide a skin area into the homogeneous regions. Finally, it extracts a valid face candidate from the segments in the area by using conditional merging with the aspect ratio of frontal face. The proposed face candidate extraction method is applied to a face detection system and tested on various color images. The experimental result shows that the face detection rate of our proposed method is higher than other face candidate extraction methods in the same face detection system.