Computer Aided Early Detection and Classification of Malignant Melanoma

摘要

The diagnosis and application of Skin Cancer using Image Processing are a non-invasive technique. Currently, a lot of methods are present in the analysis and diagnosis of lesions, which provide quantitative information regarding a lesion and act as an early-warning method for it. This presented diagnosis can be used in hospitals as an alternate method for skin cancer detection and can help domain experts in reducing the time for its classification. The proposed method focuses on the classification of Skin Cancer with high accuracy by first reducing the noise from the images using Dull-Razor software, then segmenting the image using an automatic segmentation process. Important features are then extracted from the image using the GLCM and basic statistical method. The features are then fed into SVM to classify the image data. The investigation is carried on 50 normal and 50 melanoma images obtained from DermNet and ISIC archive.