期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2013
卷号：49
期号：1
出版社：Journal of Theoretical and Applied
摘要：Compressed sensing has been widely used to reconstruct magnetic resonance(MR) images from highly under-sampled k-space data. Generally, there are two main types of dictionary: analytic dictionary and trained dictionary. In this paper, we propose a novel framework to adaptively learn the dictionary and reconstruct MR images from highly under-sampled k-space data. We use the beta-Bernoulli process as a Bayesian nonparametric prior, which can learn dictionary adaptively. This patch-based dictionary learning process can also infers the sparsity of each patch and the noise variance. Our numerical experiments demonstrate that our reconstruction is more accurate compared to existing algorithms.
