The activity image in image enhancement and segmentation
Patrick DE SMET, Danny De Vleeschauwer · Ghent University Academic Bibliography (Ghent University) · 1998
In this paper we study a scheme to segment images. The scheme consists of two steps: image enhancement based on anisotropic diffusion and watershed segmentation. In both steps an activity image plays a central role; to drive the diffusion in the enhancement step and as a topographic surface in the watershed segmentation. In this paper we extend this segmentation scheme to images in true color. Therefore, we only need to extend the definition of the activity image to images in true color. The qualitative performance of this color segmentation scheme is illustrated with results for real images in true color. 1 INTRODUCTION In this paper we study a scheme to segment images. This scheme consists of an enhancement step, based on anisotropic diffusion, and a watershed segmentation step. Central in both steps is the activity image, which is supposed to take high values in the neighborhood of edges and low values for interior pixels. So, this activity image serves as a fuzzy edge indicator. ...