Melanoma Classification Using Combination of Color and Shape Feature

摘要

According to the WHO, about 132 thousand cases of melanoma occurred each year. British Association of Dermatologists also launched, 77% of people do not recognize the symptoms of malignant skin cancer. Melanoma has a deadly effect and including one type of silent killer, but this can be early detected to be cured entirely. To recognize melanoma early, we proposed melanoma classification using Histogram of Oriented Gradients and Color Histogram. Histogram of Oriented Gradients is used to extract the shape features, while Color Histogram is used to extract the color features in HSV color space. Using Random Forest as the classifier, we obtained F1Score 93.3%, with the pixels per cell is 10x10 and cells per block is 1x1 for HOG and five bins for Color Histogram.