Published Paper Details:

Skin is an Extra-ordinary human structure. People frequently suffer from many known and unknown diseases. There are the large number of spread diseases and some of them are the most common disease in the world. The diagnosis of these diseases are very difficult because of its difficulties in skin texture, presence of hair on skin and color. Due to the lack of medical facilities available in the remote areas, patients usually ignore early symptoms which may be worsen the situation as the time progress. The diagnosis of skin Disease also take longer time. It is required to develop methods of diagnosis using machine learning in order to increase the accuracy of diagnosis for various types of skin diseases. Machine learning techniques are widely used in medical fields for diagnosis. These algorithms use feature values from images as input to make a decision. The process consists of three stages-The feature extraction stage, the training stage and the testing stage. The process makes use of machine learning technology to train itself with the various skin images. The objective of this process is to increase accuracy of skin disease detection. Three important features in image classification are texture, color, shape, and combination of these. In this work, color and texture features are used to classify the skin disease. Normal skin color is different from the skin with disease. Smoothness, coarseness, and regularity is effectively identified using texture features in the images. Hence, these two features are explored to identify skin disease effectively. In this work, entropy, variance and maximum histogram value of Hue-Saturation-Value(HSV) features are used. These features are used to build machine learning algorithm by using Decision Tree(DT)and Support Vector Machine(SVM). At first level, entropy measure is used to split the tree. At second level, variance is used to get leafs for textures. In color features, maximum histogram value of HSV measure is used to split the tree. Accuracy is used to test the performance of the proposed algorithm.