Dermatoscopy Image Segmentation Based On OTSU Algorithm Limiting Grayscale

In recent years,the incidence of skin cancer has rapidly increased.In the late skin cancer mortality rate reaches almost 100%,and the prognosis is very poor.In the early stages,if it can be diagnosed early and get timely treatment,the possibility of cure will be very high,and basically no sequelae after treatment.However,because the early symptoms are not obvious,the patient does not care,coupled with the uneven level of physician diagnosis,etc.,it is easy to delay the timing of treatment.Computer-based auxiliary diagnosis system can effectively assist physicians to improve the accuracy and efficiency of early detection and diagnosis,reduce the misdiagnosis rate and missed diagnosis rate of physicians.Skin lesion segmentation is an important step in the computer-aided diagnosis system of skin cancer and also one of the difficulties in the whole system.The higher the segmentation accuracy is,the better the result is,and the less difficult it is to extract and classify the suspicious areas of the skin at the later stage.This topic takes dermoscopy image as the segmentation object,and designs two automatic segmentation methods.Based on the OTSU algorithm with limited gray level and the smooth transition region threshold segmentation algorithm based on optimized noise reduction,we can automatically segment the image and get the complete skin lesion area.The main work of this paper is as following:1.Dermoscopy image acquisition.The experimental images were all from the radiology department of Nanyang chest hospital.The cancerous images were randomly selected and the appropriate segmentation technique was designed according to the characteristics of the images.2.Pretreatment noise reduction.Dermoscopy image contain complex information that is susceptible to noise during the segmentation process.In this paper,the most common black frame noise and hair noise in the dermoscopy image are preprocessedand noise-reduced,and the real images that are disturbed by the noise are repaired to the ideal state as much as possible so as to prepare for the segmentation in advance.3.Two types of segmentation algorithms are proposed.The first method is based on the OTSU algorithm which limits the gray scale range.After calculating the gradient image of the dermoscopy image,the algorithm limits the range of the gray value of the image to eliminate the interference of light and other noise and improves the accuracy of segmentation results.The second kind of method is based on the set of thresholding algorithm which is based on the optimization of noise reduction and smooth transition region.After extracting the smooth transition zone of the gradient image,the algorithm can obtain the target region by double-filtering the area and location of the extracted subregion and removing the unrelated sub-region.4.Post-processing the segmentation results.According to the specific results of morphological repair operation,and rely on the establishment of supporting adaptive structural elements,as well as the rough segmentation parameters to carry out the corresponding open and close operations,to solve one of the burr problems,to smooth the boundary contour,to carry out a degree of morphological expansion,in order to control the ultimate under-segmentation problem.5.Evaluate the segmentation results.Comparing the segmentation results of the two algorithms proposed in this paper with the gold standard results of the manual hand-painting,the performance of the algorithm can be judged by dividing the overlap rate and the over-segmentation rate.From the evaluation results,the OTSU algorithm proposed in this paper is better than the other algorithms in terms of the degree of overlap with the gold standard,the degree of under-segmentation and over-segmentation,the comprehensive measure and the average minimum Euclidean distance In other words,the improvement of the gray scale range is more similar to that of the gold standard.