Abstract:

BACKGROUND: Compressed sensing theory is widely used in the fast reconstruction of magnetic resonance image (MRI). According to randomly undersampling the k-space data, MRI with sparsity in the transform domain can be reconstructed exactly and this can be done by solving the constrained minimization problems using non-linear optimization algorithm.
OBJECTIVE: To enhance the sparsity of the image in transform domain and improve the quality of MRI econstruction, a new approach to weight the sparse presentation of the image is proposed in this paper.
METHODS: The nonlinear conjugate-gradient descent algorithm was utilized to solve the weighted norm minimization. In each iteration, according to the acquired image’s sparse presentation, weighted matrix was updated to enhance the sparsity of MRI.
RESULTS AND CONCLUSION: Several experiments were carried out with and without reweighting the norm. Results demonstrate that the proposed algorithm with weighted matrix can obviously improve the ability of image recovery.