Exploration into Single Image Super-Resolution via Self Similarity by Sparse Representation

Lv GUO  Yin LI  Jie YANG  Li LU  

IEICE TRANSACTIONS on Information and Systems   Vol.E93-D   No.11   pp.3144-3148
Publication Date: 2010/11/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E93.D.3144
Print ISSN: 0916-8532
Type of Manuscript: LETTER
Category: Image Recognition, Computer Vision
super resolution,  self similarity,  sparse coding,  

Full Text: PDF(7.6MB)>>
Buy this Article

A novel method for single image super resolution without any training samples is presented in the paper. By sparse representation, the method attempts to recover at each pixel its best possible resolution increase based on the self similarity of the image patches across different scale and rotation transforms. The experiments indicate that the proposed method can produce robust and competitive results.