Research On Wrinkle Detection And Quantitative Evaluation Of Face Image

Wrinkle is an important sign of aging and a focus of anti-aging,so the detection and quantitative evaluation of wrinkles is of great significance in research.However,rough skin result in more noises,and the intensity difference between the fine wrinkles and the skin background is too small,resulting in a low detection rate of wrinkles.In addition,there are problems in current quantitative evaluation methods of wrinkles,including too few evaluation indicators,lack of quantitative evaluation methods for the overall wrinkles of the face.In order to solve the above problems,wrinkle detection and quantitative evaluation methods of face images are studied in this thesis.According to the geometric constraints and intensity constraints on wrinkle curve object,the maximum curvature method is improved for wrinkle detection.Combining with the existing wrinkle evaluation methods,a quantitative evaluation method for the overall wrinkles of face is proposed.The main work of this thesis can be summarized as follows:1.Image feature extraction.In this thesis,the filtering results of Gabor filter bank and Frangi filter are combined as image features.Experiments show that the combined Gabor filter bank and Frangi filter can extract more features for wrinkles of different thicknesses.2.Wrinkle detection.An improved maximum curvature method is proposed for wrinkle detection in this thesis.The definition of wrinkles in existing wrinkle detection methods is studied,and the geometric constraints and image intensity constraints of wrinkles as curve objects are determined.After binary feature image,the effect of wrinkle feature extraction is judged according to the proportion of the eccentricity of the binary image connected component greater than 0.98.The feature image or the grayscale image is selected according to the wrinkle feature extraction effect,and the wrinkle is detected by the maximum curvature points of the image profile.Four directions are taken into account in the connection step,and a selection scheme of the connection direction of the wrinkle image is proposed to apply to the detection of wrinkles in different directions.Compared with the existing wrinkle detection methods,the proposed method greatly improves the detection rate of wrinkles,especially in rough skin images.3.Quantitative evaluation of wrinkles.In this thesis,a quantitative evaluation model of wrinkles is proposed.In view of the problem of less quantitative evaluation index of wrinkles,the medical wrinkle rating standard and the image based quantitative evaluation model of wrinkles were used to determine the quantitative evaluation indexes of wrinkles: length,depth,width and amount.According to the above-mentioned wrinkle indexes,a quantitative evaluation model of the overall wrinkle of human face is proposed.Comparing with the quantitative evaluation results with clinical scores,there is a significant positive correlation between them.