Single-image Super-resolution of brain Magnetic Resonance Images

...

Single-image Super-resolution of brain Magnetic Resonance Images Includes: - imageCollection_3DDictionaries: coupled low-resolution and high-resolution dictionary construction. Requires a set of high-resolution training images, assumed in .nii.gz format, located in a single folder and each one with a corresponding binary mask (with the same filename but ending in _mask.nii.gz). The binary mask restricts the patch extraction to a certain area within the image. Requires also specification of the number of images, the number of patches to extract per image, the low-resolution patch size (minimum [3,3,3]) and the magnification factor per dimension. - imageSuperResolution3D: super-resolution reconstruction of a low-resolution image (or a set of images). Requires a low-resolution image in .nii.gz format, with a binary mask (same filename but ending in _mask.nii.gz) and the dictionaries previously constructed. The binary mask restricts the reconstruction only to certain areas in the image, so a common interpolation is applied to the regions outside the mask. Requires also specification of the low-resolution patch size (minimum [3,3,3]), the magnification factor per dimension, the PCA coefficients percentage (useed for dimensionality reduction and speed-up) and the number of atoms in the dictionaries. - fspecial3d: 3D kernel filer construction. Requiere external packages: ba_interp3: fast 3D interpolation, available at http://cs-wwwarchiv.cs.unibas.ch/personen/amberg_brian/ba_interp3.tar.bz2 nifti: Nifti and Analyze tools, available at http://research.baycrest.org/~jimmy/NIfTI/ SparseLab: solvers for sparse optimization problems. available at http://sparselab.stanford.edu

Download