Asked by Bran
on 4 Feb 2013

I am wondering what is the best way to use the wavelet-2D decomposition package? I am trying to compress some MRI images and wondered how I know which wavelet to choose and what value of 'n' to select etc. Anyone familiar with this package who could help me with this software?

Also I was wondering if I could use soft thresholding instead of hard thresholding

Answer by Wayne King
on 4 Feb 2013

Accepted answer

In terms of soft vs. hard thresholding. Only hard thresholding would actually buy you something in terms of compression because you would end up with a large number of zeros. For soft thresholding, you shrink wavelet coefficients toward zero but you don't get rid of them, so ultimately you don't save yourself memory that way.

In terms of which wavelets to use, for image processing generally the biorthgonal wavelets 'bior#.#' are favored in practice for image processing.

Opportunities for recent engineering grads.

## 2 Comments

## Walter Roberson (view profile)

Direct link to this comment:http://www.mathworks.com/matlabcentral/answers/62207#comment_127164

It is not clear what this question has to do with compressed sensing?

## Bran (view profile)

Direct link to this comment:http://www.mathworks.com/matlabcentral/answers/62207#comment_127247

Sorry it is linked to compression. images are compressed using this software as they can be represented sparsely in the wavelet domain using some kind of thresholding. :) many thanks for your comment