TAF PUBLISHING

Image feature extraction is a step of object extraction information in an image to recognize or distinguish it from other objects. The method used for feature extraction is the Gray Level Co-Occurance Matrix (GLCM). This research is related to features calculation from the melanoma cancer and non-melanoma images using GLCM based on variations of gray level, which are 4, 8, 16, 32, and 64, and angles of GLCM orientation consisting of 4 and 8-way. The used features are angular second moment, contrast, correlation, entropy, inverse moment, and variance. Then, the feature values are used as input parameters to classify melanoma cancer by utilizing an artificial neural network (ANN). This experiment is conducted by using 45 datasets of images from www.skinvision.com. Generally, all of the experiment types’ results accurately predict melanoma and non-melanoma classification by ANN more than 93%. By inputting six parameters from GLCM feature extraction using (1) 4th degree of gray level and 4-way orientation angles and (2) 16th-degree gray level and 8-way orientation angles, we can obtain the accuracy of ANN classification by 100%.