Research Article
Single Image Superresolution Using Maximizing Self-Similarity Prior
| Input: LR image | | (1) = Bicubic(); | | (2) | | (3) Extract patches and , ; | | (4) construct training data ; | | (5) train GMM; | | (6) initialize with ; | | (7) for // represents the number of the iterative times | | Solve with (5), keeping fixed; | | Update ; | | Solve with (10), keeping fixed; | | Update ; | | end | | Output: HR reconstructed Image |
|