Shearlet-based sparse representation for super-resolution in diffusion weighted imaging (DWI)

Diffusion Weighted (DW) imaging have proven to be useful in brain architectural analyses and in research about the brain tract organization and neuronal connectivity. However, the clinical use of DW images is currently limited by a series of acquisition artifacts, such as the partial volume effect (PVE), that affect the spatial resolution, and therefore, the sensitivity of further DW imaging analysis. In this paper, a new superresolution method is presented, given the redundancy present in this kind of images. The proposed method uses local information and a multiscale Shearlet transformation to represent the directional features and the spectral content of the DW images. A comparison of this proposal with a classical image interpolation method demonstrates an improvement of about 3 dB in the PSNR measure and 4.5% in the SSIM metric.