Face Detection And Feature Extraction In Complex Background

Biometric recognition realizes personal identification based on the inherent physical or behavioral characteristics of human being, among which facial features are the most important and intuitional ones. Although the accuracy is lower than that of iris and fingerprint recognition, face recognition has become the most acceptable biometric recognition method because of its no violation of privacy and intuition. As the foundation of face recognition, face detection and facial feature extraction has become one of the most important research subject in the pattern recognition and machine visualization.It is surveyed about the methods of face detection. The technologies of face detection are usually about: knowledge-based methods, template matching methods and study-based methods. Based on these basic theories and key technologies of face detection,the papers focused on face detection and feature extraction in the complex background. The main research works and contributions as the following:(1) Analysis of the impact of light on the image, present a noise reduction algorithm based on Kalman filter and illumination compensation algorithm for image pre-processing. Experiments show that the method can obtain better results,which is good work for face detection.(2) Description and analysis of various color models, a non-linear color model is applied to skin detection and segmentation, and an improved Snake algorithm applied to human faces rapid location. This method solves Snake algorithm is sensitive to the initial position, reduced complexity, and increasing the system's real-time.(3) In order to improve the accuracy of face detection. A new face detection algorithm was proposed, based on eyes location and face features geometric match. Human's eyes have the relative rigid feature, so get the eyes location by eyes location algorithm based on Hough Transform. Then, according to the triangle relation between eyes and mouth, calculated the mouth location based on an improved geometry matching algorithm. Finally, to realized the human face location accurately and reduced the miscounts in traditional face features geometric matching algorithm.(4) In eyes location, the traditional Hough transform circle detection is need of more predictable parameters and a complex calculation. This paper presents an improved Hough transform circle detection based on the compression relation between circle and ellipse. It reduced the predictable parameters and computational complexity.(5) Through the triangle relation between eyes and mouth, calculated face rotation angle of the image. Also be applied in the dynamic image sequences to identify the best target face positive position. It is a good work for face recognition in future.