Journal
KNOWLEDGE-BASED SYSTEMS
Volume 71, Issue -, Pages 162-168Publisher
ELSEVIER
DOI: 10.1016/j.knosys.2014.07.021
Keywords
Bayesian model; Image segmentation fusion; Variational inference; Generation model; Expectation maximization
Categories
Funding
- National Natural Science Foundation of China (NSFC) [61262058, 61175047, 61170111]
Ask authors/readers for more resources
Image segmentation fusion can output a final consensus segmentation which in general is better than those of unsupervised image segmentation algorithms. In this paper, the image segmentation fusion is firstly formalized as a combinatorial optimization problem in terms of information theory. Then a Bayesian image segmentation fusion (BISF) model is proposed for a good consensus segmentation. We treat all the segmentation algorithms (or the same algorithm with different parameters) as new features and the segmentations of algorithms as values of the new features, which simplifies image segmentation fusion problems in computation complexity. Based on this idea, a generative model BISF is designed to sample the segmentation according to the discrete distribution, and the inference for BISF and the corresponding algorithm are illustrated in detail. At last, extensive empirical results demonstrate that BISF significantly outperforms other image segmentation fusion algorithms and the popular image segmentation algorithms or algorithms with different parameters in terms of popular indices. (C) 2014 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available