Change Detection Based on Conditional Random Field With Region Connection Constraints in High-Resolution Remote Sensing Images

In this paper, a novel change detection method based on conditional random field (CRF) with region connection constraints in multitemporal high-resolution remote sensing images is proposed. The change detection problem is formulated as a labeling issue to discriminate the changed class from the unchanged class in the difference image. In the CRF model, the unary potential is described by using the memberships of unsupervised fuzzy C-means clustering algorithm. The pairwise potential adopts a boundary constraint based on Euclidean distance. In addition, region iteration potential defined on a set of pixels is incorporated into CRF model to suppress the oversmooth performance. A chief advantage of our approach is to be able to achieve correct change map and avoid training a large number of model parameters. Experimental results demonstrate that the proposed method improves the change detection accuracy, is more robust against noise than other state-of-the-art approaches, and preserves boundary information.