A new approach of color images segmentation based on fusing region and edge segmentations outputs

摘要

Image segmentation requires processing of a huge volume of data. It is therefore necessary, for their implementation in industrial computer controlled systems, to search for straightforward algorithms. Presently, literature offers a lot of gray-tone image segmentation techniques, but few of them attend to color image segmentation. This paper presents a co-operative strategy within a multi-resolution color image segmentation, which attempt to extract the meaningful information (regions and boundaries), then fuse these two approaches in order to achieve an accurate, robust and suitable segmentation. A blob filling coloration algorithm allows to design, from the segmented image, a synthesis image which appears as a simplified but faithfull copy with a chosen resolution of the original image. This process is induced by approach of the human psychovisual system of perception, tender to the sharp edges, strong contrasts and large areas of color. It gives good results when applied on natural scenes, like a bunch of flowers, or on artificial scenes, like a set of building blocks. As possible, we use the coding vocabulary of R.M. Haralick.