A multi-frame image super-resolution method
Signal Processing
| Hi-index | 0.00 |
This paper addressed the issue of solving the image super-resolution reconstruction from the single or multi- frame low-resolution (LR) image to the required high resolution (HR) image. According to the LR image degradation model, we present a maximum a posteriori estimation (MAP) framework and point out that the problem of the image super-resolution reconstruction can be regarded as the inverse problem considering the minimization of the variational energy functional. Three different energy functionals including total variation integrals, Huber integrals and improved entropic integrals are considered to solve this problem. However, the above mentioned models have not local, contextual information, and thus the mosaic phenomena along image edges can not be effectively reduced. To overcome the bad phenomena, we first present a new contextual magnitude of spatial gradient definition, then propose a new class of local and contextual information driven PDE for super- resolution reconstruction. Finally, a numerical scheme and corresponding iterative algorithm is presented. The performance of all the above mentioned algorithms is compared from the PSNR and edge preservation ability points of view.