MRI Image Reconstruction Based On Compressed Sensing

Magnetic resonance imaging is called MRI, which is an important clinical monitoringtool after CT, and brings the convenience for the detection of patients with soft-tissuedisorders. However, based on MRI imaging features, during the scanning process caneasily cause motion artifacts, thus resulting in the doctors’ misdiagnosis for patients. As anemerging field of image processing theory, compressed sensing has been successfullyapplied to magnetic resonance imaging system. In this paper compressed sensing theory inMRI image reconstruction done further research, the main work is as follows:Firstly, for the shortcomings of discrete wavelet and total variation, it proposes acombination of two sparse constraints: the combination of analytical outline wave andtotal variation and the combination of discrete wavelet and higher degree total variation.under the same condition of reconstruction, the proposed two combinations compare withthe combination of discrete wavelet transform and total variation, and which has a bettereffect of image reconstruction.Secondly, according the superiority of the spare representation of the dictionary, thecombination of the global dictionary and total variation is used for MRI imagingreconstruction, and the reconstruction idea is the difference between the frequency domain.Specific implementation of the algorithm is to use alternating minimization ideas, and beconstantly updated with iterative frequency domain values, while maintaining constrantobservation sampling locations, and utimately through inverse Fourier transform to obtainhigh quality MRI images.Finally, for the shortcoming of local and global sparse, the fusion of global and localsparse for the reconstruction of MRI is proposed. In order to make the constructed imagehave more details, adaptive dictionary, dual tree complex wavelet and total variation of thethree combinations together as a sparse image reconstruction time constraints. The full useof iterative reconstruction process minimnizes ideas, alternating direction multipliermethod and difference frequency domain algorithm. Experimental results show that thealgorithm to reconstruct the image detail is more abundant and better quality.