Overview

This page provides MATLAB/C-mex code to perform most of the basic operations used in multi-frame image super-resolution. Compile the mex-files, then run SR_demo.m to find out exactly how to use the various functions.

Downloads

  • elleSR_0.2.tgz (12 KB): Super-res code bundle.
  • readme (7 KB): Readme file and description of code.

The code is offered "as is", with no guarantees and no support.

Implementation Notes

The code uses several Netlab functions, so make sure you've got that on your MATLAB path. Sparse matrix multiplication used to have memory allocation issues, so I took to using mex_amub.cpp -- you may or may not find that this is no longer necessary with more recent MATLAB releases, so see the code comments for alternatives.

The release bundle does not include more complicated functionality like the simultaneous registration super-resolution, but these can be composed from the building blocks included here with the help of the papers below.

Publications


Pickup, L. C.
Machine Learning in Multi-frame Image Super-resolution
PhD thesis from University of Oxford, 2008

Pickup, L. C., Capel, D. P., Roberts, S. J. and Zisserman, A.
Overcoming Registration Uncertainty in Image Super-Resolution: Maximize or Marginalize?
EURASIP Journal on Advances in Signal Processing (2007)

Pickup, L. C., Roberts, S. J. and Zisserman, A.
Optimizing and Learning for Super-resolution
Proceedings of the British Machine Vision Conference (2006)

Capel, D. P.
Image Mosaicing and Super-resolution
PhD thesis (2001)