Image registration of follow-up examinations in digital dermoscopy

Automated assessment tools for dermoscopy images play an important assistive role mainly due to inter- and intra-observer variations in human interpretation. Prior to classification, a robust registration module is necessary. In this paper we propose a fully automatic method for successful registration (matching) of melanocytic nevi images acquired from digital dermoscope to ensure reliable quantitative comparison of the image later. A modified SIFT algorithm (called ROI-SIFT), followed by RANSAC to identify outliers, is used to localize and match correspondence interest points, that will be used to compute affine transformations for the registration of a dermatological image with the respective follow-up examination image. Experimental evaluations of the registration method are reported for image pairs acquired during follow-up examinations. Finally, a short discussion of scheduled future work is proposed.