AUTOMATIC REGISTRATION OF IMAGES OF PIGMENTED SKIN-LESIONS

Melanoma rates in whites are increasing rapidly. Early detection, prim arily through altered appearance of pigmented skin lesions, is the str ongest determinant of survival. For high-risk patients photographic re cords are invaluable, but images must first be ''registered'' to remov e distortions introduced during imaging. This paper describes an algor ithm for registering digital images of multiple skin lesions automatic ally. The algorithm comprises four stages: (1) generate lesion maps; ( 2) identify initial matching lesions;(3) select four matching corner-l esions; and (4) transform one image. Performance tests using images of both simulated and actual skin lesions show that matching and registr ation are successful under realistic imaging conditions. (C) 1998 Patt ern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.