Region of Interest Automatic Extraction for Color Image Based on Mathematical Morphology

Combined with human visual characteristics and morphological watershed algorithm, a novel image processing method was proposed for automatic region of interest (ROI) extraction. The color image would be converted into gradient map, then smooth the map using the method of morphological opening and closing operators, last thresholding it. For discriminating the very small regions which are segmented as ROI, considering the luminance and saturation characteristics of color image, the homogeneity measure criteria was applied to merge them with their similar neighboring regions, then removed the background from the image . According to the experiment results, over-segmentation caused by watershed algorithm has been greatly improved, and this method is proved to be feasible and valid.