Name
Abstract | ||
|---|---|---|
Aiming at recovering lost detailed information in low-resolution videos, a novel video super-resolution algorithm based on nonlinear mapping and patch similarity was proposed in this paper. Taking advantage of the strong fitting ability of the convolutional neural network for the nonlinear relationship, this algorithm utilizes the learned reconstruction parameters to pad missing structures and make the low-resolution video approach to the ground truth. However, it is unavoidable to introduce irrelevant information and amplify noises in this process. The patch similarity is imported to improve the spatial-temporal consistency and anti-noise ability of this algorithm. Moreover, instead of considering the nonlocal self-similarity only, this algorithm combines the nonlocal external sparsity and spatial-temporal similarity to enrich the prior information. All these patch similarities are used to optimize the intermediate video achieved from the mapping process. Experimental results demonstrated that the proposed algorithm achieves a competitive super-resolution quality on both the subjective and the objective evaluations, when compared to other state-of-the-art algorithms. |
Year | DOI | Venue |
|---|---|---|
2016 | 10.1109/CCIS.2016.7790223 | 2016 4th International Conference on Cloud Computing and Intelligence Systems (CCIS) |
Keywords | Field | DocType |
video super-resolution reconstruction,convolutional neural network,nonlocal similarity | Computer vision,Nonlinear system,Pattern recognition,Convolutional neural network,Computer science,Ground truth,Artificial intelligence,Superresolution | Conference |
ISSN | ISBN | Citations |
2376-5933 | 978-1-5090-1257-2 | 0 |
PageRank | References | Authors |
0.34 | 11 | 5 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ling-Hui Li | 1 | 37 | 1.61 |
| Junping Du | 2 | 789 | 91.80 |
| Meiyu Liang | 3 | 18 | 8.56 |
| JangMyung Lee | 4 | 544 | 71.30 |
| Luoming Meng | 5 | 117 | 34.72 |