Software For Image Denoising Using Wavelet-Domain Hidden Markov Tree Models

Image denoising example using wavelet-domain hidden Markov tree models:
(l) Original image, (m) Noisy image (PSNR=19.9dB), (r) Denoised image (PSNR=29.5dB)

---

Code

• Self-contained MATLAB code package, including examples.
  Best suited to users who run MATLAB.
  The package contains a set of rudimentary wavelet transfrom routines that are simple but somewhat inefficient.

  Code in gzip compressed format (530kb) and PC zip format. Code for 1-D case also available in gzip compressed format.
  To unpack gzip'ed files, execute "gunzip whmt1.tar.gz" followed by "tar -xvf whmt1.zip"
  (last updated 12/7/2006).
  
  

• Self-contained C code package, including examples.
  This package uses C code for dealing with the hidden Markov tree model and runs much faster on all platforms.
  MATLAB is required only to generate the example images and to display the denoising results (see README file).

  Code in tar file format (326kb).
  To unpack, execute "tar -xvf whmtc.tar"
  (last updated 5/13/1999).
  
  

Authors:

• Hyeokho Choi, DSP Group, Rice University
  (2-D wavelet-domain HMT models)
• Justin Romberg, Georgia Tech
  (Universal HMT model for real-world images)

For further information, please see our list of wavelet-based DSP related publications.

Please report any bugs or send comments regarding these programs to jrom@ece.gatech.edu