Research Onamethod For Human Face Recognition Based On MMTD

Face recognition is one of the most challenging research topics in the fields of image processing, pattern recognition and computer vision. Existing face recognition systems can achieve high recognition accuracy under favorable conditions. While under non-ideal sampling conditions (images with low-resolution and noise), the recognition precision will fall sharply. Therefore, there are still many practice problems to be solved in the application of face recognition. In order to improve the accuracy of face recognition in complex environment, combined with the theory of high-dimensional space geometry analysis and measuring of medium truth degree (MMTD), this paper studies the method of image magnification and face recognition based on MMTD, designs and implements a face recognition system based on MMTD.Under the background of medium mathematics, based on the study of the classical image magnification techniques and MMTD-based image edge detection algorithm, this paper designs and realizes a new MMTD-based image magnification method based on the similarity of grayscale between every pixel and its neighbors in five directions. Experimental results show that the proposed method can remove the noise effectively while improve the image resolution when used to magnify the face image with low-resolution and noise, which makes the image clearer and easier to distinguish, and provides a good foundation for face recognition subsequently. Then the medium truth degree measuring criteria is studied and a new MMTD-based face recognition method is proposed. Experimental results show that the proposed method has rather high recognition accuracy. Finally we design and implement a MMTD-based face recognition system. The system uses the MMTD-based image magnification method to preprocess the face image, extracts image features using principal component analysis (PCA), and realizes face recognition with the MMTD-based criteria. Experimental results show that this system can achieve fair recognition accuracy under poor image-obtaining conditions.