Adaboost Algorithm Based On Improved LBP Features With Skin Color Segmentation For Face Detection

Face detection is originally from face recognition, which is the key section of facerecognition system. In recent years, with the high speed development of informationtechnology and intelligent human-machine interaction system promotion, theimportance of face detection in face recognition system increases obviously. Becausehuman face is inherent biological characteristics then detecting human face becomes themost potential identity authentication method. Detecting face correctly or not has asignificant impact on the subsequent recognition operations, which makes face detectiontechnology exceed the category of face recognition and become an independent researchfield, getting many researchers’ attention. This paper mainly focuses on one facedetection algorithm named AdaBoost which is a very hot algorithm in recent years. Andskin color segmentation method is used for its pretreatment operation. The maincontributions of this paper are as follows:①Several common methods for face detection are introduced and discussed in thispaper.②Skin color detection technology is introduced in detail in this paper, andseveral common color spaces are discussed, including RGB space, YCbCr space, HSVspace and so on. Then how to set up skin model in color space is discussed. Accordingto the skin color detection method based on the color space with separating brightnessinformation, a new method based on brightness information is put forward in this paper.The experiment results prove that skin color detection based on brightness informationnot only improves the detection speed and detection effect is also improved.③AdaBoost algorithm is introduced in detail in this paper and rectangle feature,integral figure and so on are also analyzed. Then how to construct weak classifier,strong classifier and cascade classifier is discussed. According to the traditional Haarfeatures which have so many features that the detection speed is slowed down, animproved LBP features are put forward in this paper. The features not only improve thedetection speed of AdaBoost algorithm, but also can describe texture features moreeffectively than primitive LBP features and multi-scale LBP features.④Because face detection based on AdaBoost algorithm is too sensitive to theimage region whose texture features are similar to human face, a modified methodcombining AdaBoost algorithm with skin color detection based on brightness information is put forward in this paper. Human face can be well detected in complexbackground image with many faces by this method, and false rate of face detection isreduced effectively.