A novel method of colour image segmentation based on Bayesian theory and edge detection

摘要

This context proposes a novel approach of colour image segmentation which combines Bayesian theory with edge detection. Firstly, the algorithm detects the image's edge and divides it into non-edge pixels and edge pixels. Secondly, some rectangles in nonedge area are drawn so that the central pixel shares the same label with all the pixels in the same rectangle. Finally, the represented pixels of the rectangles, edge pixels and scattered pixels are labeled through calculating a maximum posterior probability. As the affect of using rectangles, this method can save computing time notably and fix the number of label automatically. The experiment result shows the method is reliable and superior.