Registration-noise reduction in difference images for change detection

Abstract

Two methods are introduced and evaluated for the reduction of registration noise in difference images: image smoothing and adaptive grey-scale mapping. Landsat Thematic Mapper (TM) data obtained on two different dates over an urban and urban-fringe area depicting significant change were geometrically registered and then subtracted from each other. The TM band 3 difference image was used to test the registration noise reduction algorithms. Results indicate that adaptive grey scale mapping is more appropriate to use for registration noise reduction. © 1992 Taylor & Francis Ltd.