Automatic Detection and Classification of Skin Cancer

Cancer is a deadly disease in today’s world. Various types of cancers are spreading for which skin cancer becomes a very common cancer nowadays. Skin cancer can be of two types namely melanoma and non melanoma cancer. The objective of this paper is to detect and classify the benign and the normal image. Benign meaning the normal image, melanoma the cancerous one. And more over compare the various classification algorithms. Detection of skin cancer in earlier stages can be a life saving process. The detection of skin cancer includes four important stages namely Pre-processing, Segmentation, Feature Extraction and Classification. Detection can help in curing the cancer and hence detection plays a very vital role. In this paper, pre-processing the first and the foremost part of image processing which helps in noise removal is done by means of the median filter where the output of the median filter which is fed as an input to the histogram equalization phase of the pre-processing stage, then the input of the histogram equalized image is fed as an input to the segmentation phase where Otsu’s thresholding is done to separate the foreground and the background. The segmentation helps to identify the region of interest, Now using the area, mean, variance and standard deviation of the extracted output from the segmentation phase the calculations for feature extraction is carried and the output is fed into classifiers like Support Vector Machine (SVM), KNearest Neighbor (KNN), Decision tree(DT) and Boosted Tree(BT).Comparison of the classification is done. The algorithm shows the accuracy of the classification rate of KNN is 92.70%, SVM is 93.70%, Decision tree (DT) is 89.5% and finally the boosted tree (BT) is 84.30%.