Color Images Of Human Face Detection And Follow-up Study

The face detection and tracking technology has been widely used in many fields, such as face recognition,video conferencing, intelligent monitoring, fatigue testing, and so on. With the development of computer processing speed and deep study on computer color theory, face detection and tracking of colorized images has become a new hotspot.How to use color information effectively to improve the quality of face detection and tracking has become a subject with great significance and application value.The main research on face detection and tracking in colorized images is as follows:Firstly, an effective face detecting algorithm based on better complexion segmentation was obtained in which the YCgCr color space was adopted for its excellent complexion cluster extent and the direct least square ellipse fitting method was, for the first time, utilized in this color space. The elliptic distribution of complexion samples in CgCr subspace was acquired. For a colorized image, after the connected region analyzing, mathematical morphology processing, Euler number calculation and geometric feature of human face fitting, the face region could be screened out.Secondly,a new efficient eye-orientation algorithm, based on the color of an image and geometric feature of the eye, was proposed after deep research and many experiments. In this algorithm not only the color and brightness of the eye but also its geometric distribution on human face was taken into account.Further, in allusion to the disturbance of illumination condition and background similar to complexion in former algorithm, a new face detecting algorithm based on SVM was proposed. In this algorithm,two property indexes were obtained after two independent dimensionality reduction processing-Discrete Cosine Transform and Wavelet Transform of a gray image, by training which two separate SVM classifiers were acquired. The two classifiers were designed to be used cascadely in order to decrease wrong-detecting rate on the premise of keeping high right-detecting rate.At last,the Mean Shift tracking algorithm was improved by using Kalman Filter and self-adapting window. The tracking window was designed to be the same size as the face in the tracking process. And accurate face tracking was implemented for face moving in speed and with obstruct.