Lrtv mr image super resolution with low-rank and total variation regularizations
LRTV: MR IMAGE SUPER-RESOLUTION WITH LOW-RANK AND TOTAL
Image super-resolution (SR) aims to recover highresolutionimages from their low-
resolution counterparts forimproving image analysis and visualization. Interpolationmethods,
widely used for this purpose, often result in images withblurred edges and blocking effects. More
advanced methods suchas total variation (TV) retain edge sharpness during imagerecovery.
However, these methods only utilize information fromlocal neighborhoods, neglecting useful
information from remotevoxels. In this paper, we propose a novel image SR method
thatintegrates both local and global information for effective imagerecovery. This is achieved by,
in addition to TV, low-rankregularization that enables utilization of information throughoutthe
image. The optimization problem can be solved effectively viaalternating direction method of
multipliers (ADMM).Experiments on MR images of both adult and pediatric
subjectsdemonstrate that the proposed method enhances the details in therecovered high-
resolution images, and outperforms methods suchas the nearest-neighbor interpolation, cubic
interpolation,iterative back projection (IBP), non-local means (NLM), and TVbasedup-sampling.