| Human face detection and recognition is a major area of research within biometric signal processing and is one of the most important and most challenging tasks for the areas of computer vision and pattern recognition.Since most recognition techniques suppose the positions of facial features are located accurately, the fine location of facial features is crucial steps in most face processing system. This paper proposed a new framework to detect human faces and fine locate facial features in color images quickly. A binary skin color map is firstly obtained by applying the skin/non-skin classification algorithm, and then the integral image representation is used to preprocess the skin and intensity map. Thereafter, the refined skin areas are selected from input image for further detection. In the next processing, an intensity and features distribution model, which can be calculated with integer image very quickly, is used to decide whether the selected skin area contains a face candidate or not and where the face candidate is. In the possible face areas a new approach is designed to locate eyes accurately. Based on the positions of eyes, the other facial features are located and then the geometrical information of all facial features is used to further verify human face. In another part, this paper study the human face recognition techniques based on eigenfaces method, and the improving approach is discussed based on the results of experiments.Experiment shows the proposed method can real-time detect the frontal human faces successfully with a range of variations in color, lighting, scale and light expression from color image with complex backgrounds. The experiment on eigenfaces recognition techniques studies the merit and flaw of the recognitionmethod, and the experiment shows eigenfaces is an excellent method of image expression, but not excellent on expression of human faces. Then, this paper discusses the improving method and some new methods that may have better theory base.
|
PDF Full Text Request