Change detection in segmented images based on reference filters

In this paper we substantiate and develop a new comparative diffusion morphology technique based on filters that allow geometrical changes in the images under comparison. Two alternative interpretations of the morphological projector using “absolute” and “relative” definitions of a morphological image shape are proposed. Diffusion morphology based on superposition of two Pytiev’s projectors on mosaic shapes is considered. Diffusion morphology based on 'reference' filters is proposed as a generalization of the approach. The advantage of using diffusion morphology compared to classical Pytiev’s projective morphology in change detection applications is demonstrated.