A multispectral analysis of black skin color images for linea nigra segmentation

Linea nigra (LN) is a linear hyperpigmentation of skin which can appear in men developing prostate cancer. Early diagnosis of such cancer can be made by image characterization. There generally exist low contrast between LN and surrounding areas in black skin images that influences segmentation accuracy. In this paper, this problem is addressed through a multispectral analysis of RGB color images using Principal Component Analysis (PCA) method. This led us to find the best component that ensure low loss of significant data when converting images from color to grayscale for segmentation purpose. The relevance of the proposed approach is demonstrated by results obtained in LN segmentation.