[home] [algorithm] [results] [code] [references] [links]
Product of Edgeperts denoising
|
|
|
|
| original image | distorted PSNR 20.16 [dB] | reconstructed PSNR 31.62 [dB] |
Wavelet transforms are used as a standard preprocessing tool for image processing applications. It has been observed that wavelet transformations capture the second order dependencies between wavelet decomposition very well, i.e. the coefficients are decorrelated in the wavelet domain. However higher order dependencies remain and several statistical models have been proposed to capture these remaining statistics of the image.
Product of Edgeperts (PoEdges) are a probabilistic model to model the statistical dependencies between coefficients in wavelet decomposed images. It builds upon earlier work on wavelet statistics [1,2]. We use the PoEdges model for image denoising. So far several model have been proposed for this task an overview can be found in [3]. The state of the art in the field of image denoising is the Gaussian Scale mixture model (GSM) [4]. Our algorithm yields similar results, for a comparison have a look at the results section.
|
This material is based upon work supported by the National Science Foundation under Grant No. 0447903. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF). |
Peter Gehler http://www.kyb.mpg.de/~pgehler
Max Welling http://www.ics.uci.edu/~welling