Image noise reduction and/or image resolution improvement

A non-transitory computer readable medium encoded with computer executable instruction, which, when executed by a processor, causes the processor to:

• analyze, for each of a plurality of voxels of image data, a set of entries of a dictionary, wherein an entry represents either a mapping between a lower resolution patch of voxels and a corresponding higher resolution patch of voxel, or a local neighborhood around a voxel, and create a data matrix with rows that include patches from the dictionary that are similar to a currently processed patch around a voxel;
• derive, for each of the plurality of voxels, a subspace based on the analysis, wherein the subspace is for one of the mapping or the local neighborhood;
• and
• restore the image data to produce target image data, which has higher image resolution or reduced image noise relative to the image data, based on the subspaces, using a least-squares approach to optimize a local or global optimization.

A method for improving image quality of image data, comprising:

• analyzing, for each of a plurality of voxels of noisy image data, a set of entries of a dictionary, wherein an entry represents a local neighborhood around a voxel;
• identifying, for each of the plurality of voxels, patches in the dictionary that are similar to a currently processed patch around the voxel;
• creating a data matrix with the identified patches, wherein each row in the matrix includes a patch;
• performing a principal component analysis (PCA) on rows of the matrix to determine principal components;
• estimating, from a result of the PCA, a number of principle components, for the patches with a largest corresponding eigenvalue to model the local patch;
• deriving, for each of the plurality of voxels, a subspace based on the analysis, wherein the subspace is for one of the mapping or the local neighborhood;
• and
• restoring the image data to produce target image data, which has reduced image noise relative to the noisy image data, based the patches from the dictionary, the matrix, the principal components, the number of principle components, and a local or global optimization.