On high-resolution image estimation using low-resolution brain MRI

In the context of medical imaging, super-resolution (SR) is currently a promising post-processing technique to increase the image resolution. However, although many SR methods have been proposed in the literature, the gain of this type of approach in a real situation has not been precisely quantified. In this work, we evaluate image acquisition protocols and SR algorithms using in-vivo brain MR data as gold standard. The results show that using orthogonal image acquisition protocols lead to better reconstructed images than overlapping parallel low-resolution image stacks. Moreover, if the preprocessing steps (such as image denoising and intensity correction) are carefully performed, there is no significant differences between the evaluated SR algorithms.