Research On Automatic Face Detection Algorithm In Video Images

Firstly face detection problem is being developed which acted as localization step in automated face recognition system. In recent years, it shows important application value in many fields include that visual supervise, content-based video retrieval and new human computer interaction and so on. So many people are taking it seriously as an independent problem.This paper presents a new face detection algorithm for color video images based on skin color and multimodal information fusion.First, this paper presents a new means for selecting skin samples;and then comparing skin distribution in the eight color spaces and analyzing the adaptability for different skin patterns,poses a face initial orientation's method which uses the single Gaussian model in the TSL color spaces,and calculates skin probability images ;afterwards comprehensive comparing three typical threshold value separating algorithms ,put forwards a face separating method which bases on region growing and fuses multimodal informations;final,raises a face confirming algorithm which fuses three shape features.This algorithm that can detect multifaces with multiple orientation and scales against a complex background in many video sequences is presented, especially being the same with which face is linked to similar face region. New algorithm is applied to multiple test video sequence, and results are promising.