An unsupervised method of rough color image segmentation
Marius Tico, T. Haverinen, Pauli Kuosmanen · 2003
Image segmentation is a challenging task, and the goodness criteria depends on the target application. In an image retrieval application it is important to find those few most important objects that describe the contents of the image, and hence a rough image segmentation would be more appropriate for this task. This paper presents an unsupervised method of rough color image segmentation. The image is first segmented in the achromatic and chromatic regions and then different criteria are used to perform the segmentation inside each one of the two regions. Achieving a low sensitivity to uneven illumination conditions, the proposed technique succeeds to segment a few most prominent regions in the image. The technique can be used for the purpose of automatic extraction of region based features (shape and color) in the context of image retrieval systems.