A Novel Image Segmentation Algorithm for Color Object Image Retrieval

Image segmentation is a key problem in computer vision, it also play a very important role in many object level image retrieval system. However, automatic object segmentation is a hard work, for it is difficult for computer to determine which is object and which is not automatically. In this paper, a novel segmentation algorithm for color image was proposed, this algorithm needs ones to identify an object contained area firstly,and then a 3-D color histogram is calculated to represent the sub-image and background, at last, object can be segmented through compare the distance between object and background. This entire based on such a hypothesis:in expand of sub-image the background gray value probability distributes is continuous. This algorithm avoids the disadvantage in color image segmentation of conventional methods. Finial the experiment result is given.