A Fast Edge Tracking Algorithm for Image Segmentation Using a Simple Markov Random Field Model
Feiyue He, Zheng Tian, Xiangzeng Liu, Xifa Duan · 2012
This paper presents an fast edge tracking algorithm for reducing the computation time of unsupervised image segmentation using a simple Markov random field model (MRF). The classical two-component MRF (CMRF) based image segmentation algorithm is time-consuming for sweeping image and repeat computing all labels at each iteration process. However, most of labels remain unchanged from an iteration to the next. So most of computations are redundant and contribute nothing to the final segmentation. The proposed algorithm works by tracking edge rather than all pixels and computing their labels at each iteration. The algorithm is simple, easy to implement but fast. Experimental results show that, compare to the image segmentation algorithm based on CMRF method, the proposed algorithms can substantially reduce the computation time but the segmentation results are comparable.