Significance of Color & Texture Features in Computerized Melanoma Diagnosis Using Soft Computing Techniques

摘要

Melanoma is one of the most deadly skin cancer type. Early stages of melanoma are highly curable than in later stages. Lots of effort has been put forward the development of methods for efficient diagnosis of melanoma in skin lesions. This work addresses the importance of color and texture feature extraction for classification of melanoma from dermoscopy images, that is, the main objective of this paper is to compare the role of color and texture feature in melanoma classification. It is resulted that texture feature gains more importance than color features when they are used separately. Texture features alone give the maximum classification accuracy of 97. 5%, whereas Color features alone gives the maximum accuracy of 86%. The above classification system was tested using the dataset of 200 dermoscopy images obtained from Pedro Hispano Hospital, Portugal.