Font Size: a A A

The Research Of The Face Detection Base On The Statistical Theory

Posted on:2008-01-18Degree:MasterType:Thesis
Country:ChinaCandidate:D D YaoFull Text:PDF
GTID:2178360242971978Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Human face not only has extremely rich information but also is the key symbol for individual identification.Face information processing plays an important role in research fields of computer vision,pattern recognition and multimedia technology.As a key technology of face information processing,face detection is a hot research topics in pattern recognition and a subject of very active research that attracted universal attention in computer vision.The main research contents of the thesis are listed as followings:1)Face detection based on template sequences of sub windows in tree structure.Against the overall and local information fusion problems,the paper proposed concept that template sequences of sub windows in tree structure to expansion of the traditional template.The paper uses integral image concept to achieve rapid calculation of the correlation coefficient and support vector machines' kernel function to fusion features.The paper sharply accelerated via coarse filtration.The experiments on CMU+MIT frontal face test set result in a correct rate of 89.4%with 69 false alarms.The average processing time on a P4 1.7GHz PC is about 36s for a 364x367 pixel image by the Matlab 6.5 environment.It can be seen that the proposed method is very efficient and has much value in application.2)Face detection based on phase congruency and support vector machines.Against the effects of light conditions,The paper proposed the concept of phase congruency face.Phase congruency is a dimensionless quantity that is invariant to changes in image brightness or contrast.By the integral projection of the phase congruency image as features and the support vector machines as the classifier,The paper has been robust to various illumination and sharply accelerated via coarse filtration.Better experimental results are achieved on CMU+MIT frontal face test set and Yale Face Database B.So,this method is effectiveness.
Keywords/Search Tags:Face detection, Multiple template matching, Phase congruency, SVM, AdaBoost
PDF Full Text Request
Related items