Region-based MRF model with optimized initial regions for image segmentation
Zheng Chen, Yijun Hu, Leiguang Wang, Qianqing Qin · 2011
In this paper, we propose a region-based MRF model with optimized initial regions (RMRF-OIR) for image segmentation. In the RMRF-OIR, a modified mean shift is introduced to get the optimized initial over segmented regions. Then, a region-based MRF is used to model these initial regions on the region adjacency graph. Finally, the segmentation results will be obtained by using a region merging scheme for the region-based MRF model. Experiment results of remote sensing images prove a high accuracy and efficiency of RMRF-OIR compared with the non-optimized region-based MRF and the pixel-based MRF.