Effective Approach of Learning based Classifiers for Skin Cancer Diagnosis from Dermoscopy Images
Skin cancers encase basal cell and squamous cell and melanoma. First two are not dangerous but malignant melanoma is very dangerous and it is very difficult to treat if it goes at higher stages. Early identification of dangerous melanoma can possibly diminish destructiveness and exhaustion. Two grouping approaches for the recognition of skin cancer using learning based classifier are presented. Support vector machine and bag of visual words classifiers have been used based on Laws Texture Energy Measures to classify the skin cancer images into cancerous and non cancerous. The proposed cancer detection methods extract Laws Textures from the Malignant Melanoma and classify the suspicious regions by applying the machine learning classifier. These methods have been tested for 100 skin cancer images and from the performance analysis, the accuracy of support vector machine is 93.57% and Bag of visual words is 95.67%. This experimental result shows the performance of Bag of visual words is better than support vector machine for the recognition of melanoma disease.