Fast joint optimization in MRF-MAP-based segmentation of color images

Color image segmentation is an important topic in the image processing field. In most segmentation methods based on MRF-MAP, the segmentation quality is far behind requirement due to the absence of appearance models, and it is often suffering from the low computational efficiency of iterated calculations. To address these problems, the paper proposes a new segmentation method for color images based on MRF-MAP, which is solved approximately in an efficient non-iterated way, with the help of a slightly tuned Lanczos eigensolver. The experiments demonstrate that the new method achieves competitive performance compared with the groundtruths or other state-of-the-art methods.