Research On Face Recognition Based On Sub-image Segmentation

Face recognition is a branch of pattern recognition field, and is wildly used in people's daily life and social management. Compared with other identification methods, face recognition is more uniqueness, stability and versatility. It concerns digital image processing, computer vision, neural networks, physiology, mathematics and other subjects'content. Face regcognition is a hot field in pattern recognition.Based on the analysis of the research progress on face recognition, this dissertation has done the following works. First, combined with standard deviation and variation coefficient, a feature extraction method has proposed baed on image-segmentation idea. The standard deviation and variation coefficient is used to process the sub-images. Then, the features extracted from sub-images are combined to form the feature of the whole face image. Based on the ORL face database, BP neural network is used as the classifier to verify the performance of the feature extraction methods. The experimental results show that the method works well. the method based on variation coefficient works better if the smooth area are removed. Second, a new bionic model, KIII model, is researched. Combined with the feature extracting method based on standard deviation, KIII model shows good performance in pattern recognition. At the same time, KIII model can remember new patterns only learning the new patterns several times. Third, transductive confidence machine has been introduced to recognize different face patterns in order to resolve the problem that the general methods can not measure the reliability of predicted results. In face recognition, TCM can not only give the prediction results, but also give the reliability of the prediction results.