Faces Detection Based On Quantized Angle Features

Human face detection is an important problem in pattern recognition and image processing. And it is the key step of face recognition. With the development of intellectualized technology, human face detection has wide applications in many fields such as security access control, intelligent supervision, video conference and content-based image retrieval. Human faces are familiar to us, but theirs detection is a challenge. If we can find a high-performance solution, it will offer important enlightenment for solving other similar complicated pattern recognition problem. It makes studying fast and accurate human face detection technique have important meaning.In this thesis, we present an algorithm based on Quantized Angle features and AdaBoost to detect frontal views of human faces in sequential gray images.We make summaries, comparison, and researches on correlative technologies at the beginning of this thesis. Then we make detail analyse at face detection method base on statistical model, analyzes common face detection methods and how Quantized Angle features and AdaBoost be applied to learning and classifying.According to the concept of Quantized Angle Features in chapter 3, we make the principle to be the practical mathematical formula. Then we apply it to statistical analysis, the result show that features can be applied to pattern recognition. In chapter 4 we improve the algorithm of AdaBoost so that it can be combined with Quantized Angle Features. In chapter 5, we introduce the realization of face detection and show the result, and show the experiment result of scaling method and corresponding parameter in detection system. The experiment result shows that the algorithm works efficiently in the gray images of complex scenes.At last, we carry on comprehensive analysis to the operation result of system and point out the further work.