Introduction to Super Resolution from an Expert
There are three major groups studying image super-resolution in the world.
- The first is Galatsanos’s group http://www.cs.uoi.gr/~galatsanos/Journal%20Papers_TOPIC.htm
- The second is Milanfar’s group http://users.soe.ucsc.edu/~milanfar/publications/
- and the last is Katsaggelos’s group http://ivpl.eecs.northwestern.edu/people/akatsaggelos
The super-resolution is quite close to the image restoration/deconvolution. So, it’ll be helpful to check such relevant papers. The above three groups has published many papers, but heuristics are often included in their papers. I guess it is better to know a rather simple basic theoretical foundation.
From that point, I will recommend to see Bishop’s paper at first. Bishop is a pioneer of the super-resolution from Bayesian stand point: Bayesian image super-resolution Tipping, M. E., & Bishop, C. M. (2003). is a paper presented at NIPS.
We (people in Ishii Lab) also invented a Bayesian super-resolution by incorporating line process originally invented by Geman & Geman. Superresolution with compound Markov random fields via variational EM algorithm
For a practical use, it is important to process quickly. Okutomi’s lab has invented good algorithms that can be used in practice. Actually, they sell their system for commercial use now. Their papers are worth reading.
- A Fast Algorithm for Reconstruction-Based Superresolution and Evaluation of Its Accuracy Masayuki Tanaka and Masatoshi Okutomi Systems and Computers in Japan, Vol.38, No.7, pp.44-52, June, 2007
- A Fast MAP-Based Superresolution Algorithm for General Motion M.Tanaka and M.Okutomi Proc. of SPIE-IS&T Electronic Imaging 2006, SPIE Vol. 6065-49, pp. 1-12, 2006.
- Robust and Accurate Estimation of Multiple Motions for Whole-Image Super-Resolution Masayuki Tanaka, Youichi Yaguchi and Masatoshi Okutomi Proceedings of 15th IEEE International Conference on Image Processing (ICIP2008), pp.649-652, October, 2008
Source: Dr. Shin-ichi Maeda