An Analysis On Human Skin Texture And Aging Evaluation

With the improvement of the living standard,facial beauty and skin care have been widespread concerned.Therefore, how to evaluate current condition of facial skin objectively and quantitatively has been a research focus regarded in image processing in recent years. It has broad applying value in practice. In this thesis, depending on previous studies,and combined with mathematical algorithms and computer simulation based on network, a kind of image texture tatistical feature analysis and recognition system was explored.Firstly, the problems of current research and future trends were presented, including the spatial gray level co-occurrence matrix method, fractal dimension and Gabor wavelet transform method. These types of statistical methods in the meaning of the various features of the actual texture were analyzed. By comparing the advantages and disadvantages of various methods the spatial extraction of GLCM texture features was employed.Secondly, an effective method to enhance the image texture information was put forward by analyzing and comparing existing methods of image pre-processing means. Then after researching the traditional arithmetic of BP network, a new improved TFBP (Transfer Function Error Back Propagation) network arithmetic was adopted as an artificial neural net pattern recognizing arithmetic. That is the transfer function of the error back propagation arithmetic for feedforward neural networks.Finally, during the training experiment of skin texture image, the training has been underway to TFBP network with the neural network toolbox of MATLAB. Then the face skin texture images are simulated to identify. The result shown that classification recognition was correctly lead to 87.5%. In this thesis, the neural network arithmetic based on TFBP face skin texture feature realized the quick classification and identification. The experiment results prove that the measuring system is feasible.