A Generalized Two-level Bregman Method With Dictionary Updating For Non-con Vex MRI

Magnetic resonance imaging(MRI) as a new medical imaging diagnostic technology is developing very rapidly in recent years. MRI has no radiation and ionization effects, and high contrast resolution of soft tissue. At the same time, it has boneless artifact interference. But there exists a deadly problem that is scanning for a long time. It is not suitable for the inspection of the movement and critically ill patients, especially for patients with noise or loss of self-control, without the use of sedatives, it is difficult to imaging. So how to reduce the scanning time and rechieve better reconstruction images becomes a critical problem.In order to solve this problem, this paper proposes a new dictionary learning algorithm named generalized Two-level Bregman Method with Dictionary Updating(GTBMDU). Using thepl norm to replace1 l norm, the generalized thresholding minimizes the non-convex p-norm based function with p ?1, then alternately updates dictionary and image block. It can produce more precise and effective sparse representation, thus to reduce the resonance calculation sample rate, and reduce the scanning time at the same time, can also ensure the quality of image reconstruction.Because GTBMDU algorithms use the modified thresholding to minimize non-convex functions, that penalizes small coefficients over a wider range and applies less bias to the larger coefficients. So under the framework of the algorithm, it can achieve improving the scanning speed and realizing high quality imaging of sampling data. In order to assess the effectiveness of the proposed method, it has done a lot of simulation experiments. The experiments show that Relative to TBMDG, the peak signal-to-noise ratio of the GTBMDU algorithm improves 0.7~1.2 dB.