Melanoma is a malignant skin tumour. If detected and surgically removed early whilst residing in the superficial part of the skin the prognosis is excellent. A seven-point check-list of signs and symptoms has been adopted by the Cancer Research Campaign to help non-dermatologists distinguish benign pigmented lesions from melanoma. The presence of irregularity in shape or outline of a mole is one of these important signs. However, it has recently been shown that not only patients, but also clinicians have difficulty in agreeing upon whether a mole exhibits irregularity or not. Computer image analysis methods have been developed to derive quantitative measures of those shape parameters which dermatologists appear to use in their assessment of shape irregularity. The overall shape of the lesion is expressed by the 'bulkiness' measure. Irregularity of the border is expressed by two fractal dimension measures, one for the 'structural' aspect of the shape and the other for the 'textural' aspect. These measures were used in combination to classify melanomas in the study containing silhouettes of 43 melanomas and 45 benign lesions producing correct classification with 91% sensitivity and 69% specificity. This paper describes computer image analysis aspects of the study.