- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi experts,
Is there functionality available within IPP to perform translation invariant wavelet transforms (also
known as non-critically sampled WT)? Alternatively, if someone can advise me on how to
build one, based on the WT components already available, that would be great too..
Thanks,
Sid.
Is there functionality available within IPP to perform translation invariant wavelet transforms (also
known as non-critically sampled WT)? Alternatively, if someone can advise me on how to
build one, based on the WT components already available, that would be great too..
Thanks,
Sid.
Link Copied
3 Replies
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Sid,
As you see, IPP provide tranditional DWT, see all WT function lists provided by IPP and schematic diagramin IPP manual:
C:\Program Files (x86)\Intel\ComposerXE-2011\Documentation\en_US\ipp\ipp_manual\index.htm
and the resampling stepswas capsulated in the function.
Butdo you have exact relationship conversion between invariant wavelet transforms and DWT? if yes, thenyou may consider toimplement one based on the existingWT.
Regards,
Ying
As you see, IPP provide tranditional DWT, see all WT function lists provided by IPP and schematic diagramin IPP manual:
C:\Program Files (x86)\Intel\ComposerXE-2011\Documentation\en_US\ipp\ipp_manual\index.htm
and the resampling stepswas capsulated in the function.
Butdo you have exact relationship conversion between invariant wavelet transforms and DWT? if yes, thenyou may consider toimplement one based on the existingWT.
Regards,
Ying
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Ying,
Thanks for the reply. I must say, I was trying to avoid implementing, but it seems that
there is no other alternative. non-critically sampled wavelets zero pad the filters before convolution at each level, hence at each scale we get 4 bands of the same size as the original image. A lot of things change because of this, but in the implementational sense, this is the basic difference, and the advantage is that point-to point correspondence is maintained which is beneficial for noise removal applications.
With this information, maybe you can giude me a bit closer to the existing functionality in IPP for a working
solution.
Thanks again,
sid.
Thanks for the reply. I must say, I was trying to avoid implementing, but it seems that
there is no other alternative. non-critically sampled wavelets zero pad the filters before convolution at each level, hence at each scale we get 4 bands of the same size as the original image. A lot of things change because of this, but in the implementational sense, this is the basic difference, and the advantage is that point-to point correspondence is maintained which is beneficial for noise removal applications.
With this information, maybe you can giude me a bit closer to the existing functionality in IPP for a working
solution.
Thanks again,
sid.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Sid,
Sorry for missing your reply.
Here is C codeusing Intel IPP WT.
Code Samples for the Intel Integrated Performance Primitives (Intel IPP) Library 7.0*
w_ipp-samples_p_7.0.7.064\ipp-samples\image-processing\2d-wavelet-transform
as the manual, it require the src is 2xdst. So itseemsnot workable to pad the filters to get same size of original image.
Best Regards,
Ying
Sorry for missing your reply.
Here is C codeusing Intel IPP WT.
Code Samples for the Intel Integrated Performance Primitives (Intel IPP) Library 7.0*
w_ipp-samples_p_7.0.7.064\ipp-samples\image-processing\2d-wavelet-transform
as the manual, it require the src is 2xdst. So itseemsnot workable to pad the filters to get same size of original image.
Best Regards,
Ying

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page