Computational Intelligence for the identification of skin melanoma

The main objective of using computational intelligence is to classify the skin cancer images into three classes as Common nevus, Atypical nevus and Melanoma using a faster and accurate multi-classification technique. This work proposes color local directional pattern (CLDP) for feature extraction. This descriptor combines the vital features for skin melanoma like color, texture and shape into one vector and uses Dull-Razor algorithm for hair removal in the pre-processing step. The proposed model uses Extreme Learning Machine (ELM) for a faster and accurate multi-classification than support vector machine (SVM). A classification accuracy of 95.23% is achieved on PH2 dataset.