Face Detection Based On AdaBoost

Face detection has become an important research theme in the topic of Pattern Recognition and Computer Vision in recent years. The research of face detection is valuable in both theoretical and practical so it attracts great attention from many researchers. Face detection has a broad application in many fields such as human-computer interface (HCI), securities, access con-trol, video surveillance and content-based image and video retrieval. Thanks to the hard-working of many researchers that hundreds of different face detec-tion algorithms have been proposed in many papers. In 2001, Viola and Jones proposed a fast face detection algorithm based on weak features and AdaBoost learning algorithm. It is considered a mile-stone in this field. Weak features, AdaBoost learning algorithm and cascade classifier have been introduced into this field. It’s the first time that a face detector works on real-time face detection and this makes face detection practical.This article describes the Viola and Jones’s face detector and the theo-ry of AdaBoost learning algorithms. The rapid face detection algorithm has already been widely known, however it’s hard to get a well performed face de-tector due to only very few references that on the discussion of training face detector can be found. In this paper, the author has tried many methods to train a well performed face detector and some suggestions are given through theoretical analysis and experimental results. Significant improvements of the performance for face detector can be made by following the suggestions in the training procedure.The author has also made a survey to the recent advantages in the past decade and crucial progress is introduced in this paper. The progress includes two parts. One is feature selection such as LBP and the other is more sophisti- cated AdaBoost learning algorithms. Experimental results on these new tech-nologies are given by objective performance evaluation for further references.