Single-Image Super-Resolution via Low-Rank Matrix Recovery and Joint Learning
Xiao Chen · Chinese Journal of Computers · 2014
This article proposes a novel single-image super-resolution reconstruction method based on the low-rank matrix recovery and the joint learning technique.First,the training samples are divided into subsets by patch similarity.Second,the low-rank matrix recovery is utilized to learn the underlying structure of each subset.Then the joint learning technique is employed to train simultaneously two projection matrices,which map the low-rank components of original high-resolution and low-resolution features onto a unified space.Finally,the neighbor embedding-based image super-resolution reconstruction is performed in this unified space.Experimental results suggest that our method outperforms several super-resolution approaches both quantitatively and perceptually.