Some medical applications of example-based super-resolution

Example-based super-resolution (EBSR) [3, 4] reconstructs a highresolution image from a low-resolution image, given a training set of high-resolution images. In this note I propose some applications of EBSR to medical imaging. A particular interesting application, which I call x-ray voxelization, approximates the result of a CT scan from an x-ray image. 1 Example-based super-resolution Example-based super-resolution (EBSR) [3, 4] produces a high-resolution image from a low resolution image, given a training database of high-resolution images. The basic idea is that an observed low-resolution pixel is the average of a set of high-resolution pixels. The training set can be viewed as a probabilistic map from a set of underlying high-resolution pixels to the corresponding observed low-resolution pixel, since in general it is straightforward to model the blurring and subsampling process that turns a high-resolution image into a low-resolution one. Naively EBSR could be solved using a nearest neighbors scheme, but of course it is necessary to make use of inference and machine learning techniques to handle the spatial relationships between pixels (see [3, 4] for details). 2 Applications to medical imaging While super-resolution has many application, EBSR for medical applications seems to have been under-explored (see [6] for a rare example). Yet there are many medical applications where EBSR would be extremely natural.