Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- Intel Community
- Software
- Software Development SDKs and Libraries
- Intel® Integrated Performance Primitives
- How to Implement Multilevel Wavelet Forward/Inverse Transform

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

Kadir_Kirtac

Beginner

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

12-03-2010
04:12 AM

67 Views

How to Implement Multilevel Wavelet Forward/Inverse Transform

The availableWTFwd andWTInv routines perform single level wavelet transform and its inverse.

Is there any available code to perform multi level transform?

I tried to implement manually but failed. The forward transform results seem ok, but reconstruction results oher than the one of the smallest scale seem totally wrong.

What I am doing is, I am just applying the single level inverse transform routine at each level(scale), using the previous reconstruction result as the approximation image of that level and using the current level detail images(detail coefficients HL, LH and HH of the current level) without any change. So that is the standard multi level wavelet reconstruction procedure, as far as I know.

With the same implementation I mentioned above, I can successfully apply single level forward and inverse transforms, but when I change the level to a number that is greater than 1, my reconstruction results become totally wrong, taking the form of a periodic pattern. But smallest scale approximation band which I obtain by 4-level fwd transform and my reconstruction result obtained by the multilevel inv transform at the smallest scale become the same. But as I am going up during the reconstruction(bigger resolution scales) , my reconstruction result goes worse.

I enclose the input image, approximation band of the single level wavelet transform(with padding), single level inverse transform result(reconstruction, padding removed), and 4-level reconstruction result. Notice that the 4-level reconstruction result makes no sense compared with the input image..

What can be the cause of my erroneous result?

Link Copied

3 Replies

Vladimir_Dudnik

Employee

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

12-03-2010
04:22 AM

67 Views

did you noticed 2d-wavelet-transform sample in IPP sample package (image-processing folder)?

I also would recommend you to look at UIC JPEG2000 codec sample where multi-level WT implemented.

Regards,

Vladimir

Kadir_Kirtac

Beginner

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

12-03-2010
04:53 AM

67 Views

Quoting Vladimir Dudnik (Intel)

did you noticed 2d-wavelet-transform sample in IPP sample package (image-processing folder)?

I also would recommend you to look at UIC JPEG2000 codec sample where multi-level WT implemented.

Regards,

Vladimir

Hi,

the iplwt implementation(under "ipl-ipp" directory) and the wavelet sample under "2d-wavelet-transform" directoryonly support one-level wavelet transform.

could you tell me where exactly the multi-level wavelet code is?ipp-samples\image-codecs\uic\src path contains 6 more directories and many cpp and h files exist inside each of them.

Kadir_Kirtac

Beginner

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

12-03-2010
06:47 AM

67 Views

any comments about my first message?

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

For more complete information about compiler optimizations, see our Optimization Notice.