Improvement On Fast Face Detection Algorithm And Its Application To Video

Face detection can be applied to face recognition,human-computer interaction, visual monitoring,image retrieval based on content and so on.It attracted many researchers' attention,leading to a variety of detection algorithms However,most of them are slow,which prevents them from application in the scenarios with real-time requirement.This problem has been solved by Viola and Jones,who proposed the famous Boosted Cascade algorithm.This paper tries to improve the original method in two aspects.One is to replace the original Unique Threshold Weak Classifier with the Binary Threshold Weak Classifier.This improvement has successfully reduced the training time a lot and increased the detection speed a little bit.The other one is to use half-face training samples rather than whole-face for training,which is endowed with the ability to detect faces with occlusion,illumination and rotation.In addition,a video face detection is implemented to extend the proposed face detection algorithms from still images to moving videos.Its detection speed can almost meet the real time requirement.