Segmentation And Feature Extraction Of Infantile Hemangioma Image

With the combination of information technology and clinical medicine becoming more popular,image-assisted diagnosis technology has increasingly become a hot spot for scientific research and practical application.This subject aims at research the key technology of the infant hemangiomas image,which are obtained from ordinary camera devices(mobile phones,cameras,etc.),assistant diagnosis.This work mainly involved in the segmentation of lesions in infant hemangiomas images,image-based disease duration quantification methods,and hemangiomas image fusion.The main results of this thesis are as follows:(1)The preprocessing and segmentation of infantile hemangiomas images are studied and a Markov random field segmentation method based on hybrid features is proposed in the second chapter.Using median filtering,limiting contrast histogram equalization,and lighting constancy based on color constancy algorithm,image preprocessing;(2)The method of quantifying disease duration based on image features was studied.This study compares the extraction of image color features and texture features,and the correlation of the quantification of disease duration.The experiment verifies the effectiveness of the disease duration quantization method based on color features.(3)The registration and fusion of infantile hemangiomas images were studied.The registration method based on maximum mutual information,which was compare the SIFT registration method,was used to realize the registration of two images.Quasi-image fusion method based on wavelet decomposition is adopted to realize the fusion of image lesions.(4)The problem of sample expansion of infantile hemangiomas images was studied,and a method based on hemangiomas variation rule and fusion images for sample expansion was proposed.Experiments show that Markov random field segmentation based on mixed features can effectively segment the lesion area of infantile hemangiomas.The quantification of the disease duration of hemangiomas based on image features can assist physicians and patients in scientific quantification of disease progression and healing processes.Finally,the method based on the variation of hemangiomas and the fusion of images to expand the sample data can be validated by experiments to effectively expand the sample data and improve the accuracy of classification.