Abstract
The skin, our body's largest organ, shields us entirely. Among cancers, skin cancer, mainly prompted by sensitivity to sunlight's ultraviolet rays, is particularly fearsome. Melanoma stands out as the most perilous, originating in diverse ways. Detecting skin cancer early proves challenging for patients. Literature suggests the utilization of both manual and automated deep learning features in diagnosing skin cancer through traditional and deep learning methods. This study compares skin cancer diagnosis techniques, focusing on handcrafted and non-handcrafted features. Clinical features like the Menzies method, seven-point detection, asymmetry, border color and diameter, visual textures (GRC), local binary patterns, Gabor filters, random fields of Markov, fractal dimension, and oriental histography are investigated in the detection process. Parameters like the Jacquard index, accuracy, dice efficiency, preciseness, sensitivity, and specificity are assessed on benchmark datasets to compare techniques. The paper concludes by describing publicly available skin cancer datasets and highlighting remaining issues.