Pattern fed objects: a new approach for textured image segmentation

Abstract

Pattern Fed Objects (PAFOs) are software objects devoted to image segmentation according to texture homogeneity. They live in the two-dimensional world of images with the goals of surviving and proliferating as microorganisms. A PAFO is made of a chromosome in which texture parameter values that reflect the PAFO's relish for learned textures are coded. During its youth a PAFO is fed with different textures belonging to a coherent set, and is taught to recognize the characteristic parameters of this set. To segment an image having an unknown zone distribution, various PAFOs are spread over the image and allowed to compete. Each PAFO springs up on regions of the image as far as the underlying texture is an acceptable regimen. Some generations later, segmentation is achieved. The basic concepts of the proposed method are detailed. Our first results dealing with artificial textured images are shown and discussed.<>