Face Detection Using Skin Likelihood and Haar Features for Digital Video Processing

Abstract

Face detection is an important early step in many computer vision systems. By using pixel-wise detectors, spatial analysis of skin probability and skin regions segmentation, a new method for face detection is introduced. In this project, we proposed and implemented a modified self-organizing mixture network (SOMN) which specifies the distribution of objects in image and skin and non skin color model, skin likely-hood to exactly identify skin region of interest from image. Bayesian Decision Rule is applied to specify c as skin color or non skin color. Finally, we are using haar like features to identify face and cascade to improve performance and efficiency. We present results of an extensive experimental study which clearly indicate high competitiveness of the proposed method and its relevance to gesture recognition