Abstract | ||
---|---|---|
To alleviate the blurring effect of super-resolved faces in super-resolution (SR) field, a number of sparse representation methods have been proposed recently. However, current researches normally treat all facial patches equally and do not consider the fact that different facial patches may have unequal contributions to the SR. Regarding that AdaBoost, a classical ensemble method, has a natural weighted update scheme, this paper aims to develop a weighted-patch super-resolution approach based on the framework of AdaBoost. In the training phase, each facial patch is weighted automatically according to the difference between reconstructed patch and original patch in current iteration, which can assign more weights to the worse performed patches with lower reconstruction power in next iteration. The experimental results on two benchmarks demonstrate the effectiveness of the proposed approach. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/ICMLC.2018.8526924 | 2018 International Conference on Machine Learning and Cybernetics (ICMLC) |
Keywords | Field | DocType |
Statistical learning,Face image super-resolution,AdaBoost | Iterative reconstruction,AdaBoost,Pattern recognition,Computer science,Sparse approximation,Artificial intelligence,Image resolution,Superresolution,Encoding (memory) | Conference |
Volume | ISSN | ISBN |
1 | 2160-133X | 978-1-5386-5215-2 |
Citations | PageRank | References |
1 | 0.35 | 0 |
Authors | ||
5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shanjun Mao | 1 | 1 | 0.35 |
da zhou | 2 | 4 | 1.30 |
Yiping Zhang | 3 | 1 | 0.69 |
Zhihong Zhang | 4 | 100 | 15.85 |
Jing-Jing Cao | 5 | 1 | 0.69 |