Face Detection Based on Edge and Intensity Information

Abstract

This paper presents a new approach to locate human faces in a complex background, even when the number, sizes , and locations of faces are unknown. Furthermore, the proposed approach avoids both the overhead of explicitly construction a priori face model and the full scans of the images with all possible sizes and locations of human faces. The method is based on edge and intensity information and built on hypothesis generation-and-verification paradigm. First, and automatic segmentation method is proposed to extract all the darker pixels exact to eyes. The any two eye blocks can be used to form a face region. On the other hand, the face sides and cheeks are extracted form edge information and used for hypothesis generation and verification. In addition to these type of constraints related with face organs, the spatial disposal of face organs and the spatial relations of face patterns are also used to find out authentic human faces. Experimental results prove the practicality and feasibility of the proposed approach.