Computer aided lesion border detection applied to macroscopic images
Martin Metz, W.-D. Groch, John S. H. Curnow, Emmanuel Ifeachor, Peter J Kersey · 2001
Computer based techniques are playing increasingly important role in the diagnosis and early detection of malignant melanomas. However, this requires classification of pigmented lesions into benign and malign using imaging techniques and this is a challenging task. For the classification process first meaningful features must be extracted from the digital image. For feature extraction and the further classification process it is important to find the accurate boundary location of the lesion. There are many well known image processing algorithms, but there are also many problems to solve to get the best fitting mole border. In this article the main image processing problems for lesion border detection are discussed. Subsequently, two different approaches, namely edge and region based approaches to detect the lesion boundary, are described. We present example solutions, including an interactive region based method, which is mainly based on distance measurement in the CIE L*a*b colour-space and global thresholding of the resulting distance map.