Image segmentation and representation based on Voronoi diagrams
Yang Wang · Jounal of Xidian University · 2000
This paper proposes an image segmentation and representation method based on Voronoi diagrams, which is composed of a segmentation process and a refinement process. An image is firstly segmented into approximately homogeneous polygon regions by a split and merge process, and then a refinement procedure is conducted, which minimizes the variation in gray level values within each Voronoi region by adjusting the position of its seed, so that boundaries between adjacent Voronoi regions can run parallel to the contours of the image. Mean gray level values within Voronoi regions are also utilized to interpolate the original image and a smoothing filter is introduced to further reduce the description error on region borders.