Deconv LR method - Restoring pixel borders (reducing ringin effect)

Deconv LR method - Restoring pixel borders (reducing ringin effect)


I tried the LR deconv method from IPP and got it working. However, I am not sure as to what I should use to restore the pixel border before calling this function ? I use edgetaper in matlab for this purpose otherwise I get the ringing effect on my deblurred result. 

Kindly help with some code snippets that demonstrates how this ringign effect is taken care in IPP ? 



5 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

>>... that demonstrates how this ringign effect...

How that Ringing Effect looks like ( before it was restored )? Could you attach screenshots for Uncorrected and Corrected images? Thanks in advance.

Since IPP mostly requires you to add you own border pixels, since the kernel overlaps the image edges, then you must allocate a temp image buffer and then copy over your source image using for instance ippiCopyReplicateBorder. This will put you image in the middle and then duplicate pixel values in the border. When filtering, the source ROI should be middle part of the image (your source image). The destination ROI is an image with he same dimension as your source image. The duplicated pixels will reduce ringing.

Thanks Thomas. That works and reduced ringing to quite an extent !


Your'e welcome. I'm happy my knowledge was useful for you.

Like you, I'd very much like if Intel IPP would add a training class area to the IPP web site.
This area could be read-only and could be like a forum with topics and articles. One topic could be "Reducing ringing effects when filtering". Another could be "Implementing digital filters in the frequency domain" etc.

Leave a Comment

Please sign in to add a comment. Not a member? Join today