Intel® Integrated Performance Primitives
Deliberate problems developing high-performance vision, signal, security, and storage applications.

2D MultiLevel Wavelet Question

lee__brian
Beginner
386 Views

Hi,

I'm trying to correctly implement the Wavelet Decomposition/Reconstruction and I'm having trouble getting it working for 2D images with 'n' levels of decomposition due to the lack of documentation you guys provide. I'm able to successfully implement the Wavelet multilevel functionality for 1D arrays, but I am unsure what the correct approach is when it comes to images. I'll list the questions I have below

  • When we apply a Forward Transformation, we get the approximate, horizontal, vertical, and diagonal details.
    • If we wanted to apply another Wavelet decomposition, what exactly constitutes the High pass or Low pass from the data above?  Do we do some sort of convolution or some transformation from the Horizontal and Approximate data to get the coarse data so that we can decompose the signal one step further? Or do we just pass in the Horizontal data as the input?
  • How exactly are the Low/High anchors setup when dealing with 2D multilevel wavelets?
    • How do we exactly synchronize the group delay? Attempting to modify the anchors for the low/high pTaps to be used to create the image border results in the Forward Wavelet spec structure and work buffer not being instantiated correctly. 

These questions may be answered in your code sample documentation, but I do not have access to any 2D multilevel sample code you guys provide because I do not have the license to any of the legacy IPP libraries.

 

Thanks,

-Brian

0 Kudos
0 Replies
Reply