A total variation based approach to correcting surface coil magnetic resonance images

摘要: Abstract Magnetic resonance images which are corrupted by noise and smooth modulations corrected using a variational formulation incorporating total variation like penalty for the image high order modulation. The optimality system is derived numerically discretized. cost functional used non-convex, but it possesses bilinear structure allows ambiguity among solutions to be resolved technically regularization practically normalizing maximum value of Since convex in each single argument, analysis formulate condition terms primal–dual system. To solve system, nonlinear Gauss–Seidel outer iteration minimized with respect one variable after other an inner generalized Newton iteration. Favorable computational results shown artificial phantoms as well realistic magnetic images. Reported times demonstrate feasibility approach practice.