Variational bayesian super-resolution reconstruction

In many imaging applications, acquiring an image of a scene with high spatial resolution is not possible due to a number of theoretical and practical limitations. These limitations include for instance the sensor resolution, the Rayleigh resolution limit, the increased cost, data transfer rate, and the amount of shot noise due to the size of the digital sensor, among others. In these cases, super-resolution (SR) methods can be utilized to process one or more low-resolution (LR) images of the scene together to obtain a high-resolution (HR) image. The basic principle of super-resolution is that changes in the LR images caused by the blur and the (camera or scene) motion provide additional data that can be utilized to reconstruct the HR image from the set of LR observations. Super-resolution methods are widely utilized in a number of imaging fields, such as surveillance, remote sensing, medical and nano-imaging.