Performance Enhancement of Skin Cancer Classification Using Computer Vision

Abstract:

Skin cancer is one of the most common and life-threatening diseases, where early diagnosis is essential for effective treatment. This work presents a computer vision–based approach for enhancing skin cancer classification using image processing and machine learning techniques. Dermoscopic images are preprocessed through resizing to 250×250 pixels, median filtering for noise removal, morphological operations for hair removal, and lesion cropping. Gray-Level Co-occurrence Matrix (GLCM) features such as contrast, correlation, homogeneity, energy, entropy, mean, variance, skewness, and kurtosis are extracted from grayscale images to characterize lesion texture. A K-Nearest Neighbor (KNN) classifier with Euclidean distance is applied for binary classification of skin lesions. The dataset is divided using an 80–20 train-test split for performance evaluation. The proposed method effectively classifies skin lesions into benign and malignant categories. The results indicate that the framework provides reliable performance with low computational complexity, making it suitable for computer-aided skin cancer diagnosis systems.

Keywords: Skin Cancer Classification, Computer Vision, Image Processing, GLCM Features, K-Nearest Neighbor (KNN), Dermoscopic Images, Texture Analysis, Benign and Malignant Lesions.