Image reconstruction in MRI: regularized approach by markov random fields

Here, we address the problem of improving the trade-off between acquisition time and image quality in magnetic resonance imaging (MRI). We propose an original method based upon an inverse problem approach to image reconstruction in MRI. The image is modeled as a nonquadratic convex-potential Markov random field and a half-quadratic minimization technique is used to compute the solution. Advantage is taken of the specific structure of the MRI reconstruction problem so as to improve the numerical efficiency of the method. Tests performed on synthetic and real data show a satisfactory behavior of the proposed method.