Segmentation For Qusic-circular Accumulation Object With Multiplex Method

Image segmentation for quasi-circular accumulation, with its enormous research value, becomes one hotspot of present image division research. This paper makes a research on this image segmentation method in a multi-method inosculating way.First, a double-gliding-window method is put forward, in which the small window is the basic unit of segmentation, while the big window is basic unit of texture information statistics. Every small window is segmented according to the information statistics of big window. The experiment result shows the contradiction between the precision and accuracy of texture statistics can be resolved with this method.Secondly, a scanning segmentation method is presented here. The image is divided into N parts with the image center as basic point, and then the accumulation area edge is scanned. The N scanned edge points are connected into a polygon. The final segmentation area comes out after filling this polygon. It can be concluded that a better segmentation edge can be got with this segmentation method.Thirdly, the advantage and disadvantage of the gray co-occurrence matrixes algorithm and the scanning method are analyzed and inosculated. The scanning method is used in the area of Gray co-occurrence matrixes. It's shown that this inosculating way can resolve the problems caused by that the Gray co-occurrence matrixes algorithm has excessive segmentation but the scanning method has no one nearby the edge.Finally, this paper get an evaluation standard, showing the higher accuracy the segmentation result are, the better the edge anastomoses has.