| Magnetic resonance image is an imaging technique developed in the 1980’s.At present,magnetic resonance imaging has been widely used in clinical medicine and other fields.However,MRI has some disadvantages such as slow imaging speed,artifacts and so on.The hot topic of MRI is to reduce the scanning time while improving the resolution of the image.On the basis of the theory of magnetic resonance imaging and the image reconstruction algorithms utilizing compressed sensing,this paper mainly studied the following two problems:First,based on the non-local similarity of MR images,we studied the image reconstruction algorithms with both sparsity and low rank property.The classical algorithm,called CS-CG,uses the sparse property of the image under Fourier and total variation domain to reconstruct the image.Because the matrix made up of some image blocks with similarity is low rank,this paper improved the CS-CG method by combining the low rank property and sparsity to reconstruct MR image.The experimental results show that the improved model outperforms the other several representative methods for MRI in terms of quantitative metrics and visual effect.Second,we can make full use of the redundant data,which lie in the dynamic magnetic resonance image(DMRI),especially the redundant data on the time axis,to improve the reconstruction efficiency.This paper introduced the reconstruction model using the low-rank and sparse partial separation algorithm based on compressed sensing.This model assumes that the DMRI is a liner superposition of the background elements and dynamic elements.The background elements are low-rank and dynamic elements are sparse.In this paper,we improved the low rank partial algorithm to solve the model.The experimental results show that the improved algorithm can speed up the running time and reduce the computational complexity. |
