Bug 147648

Summary:	Make clone tool aware of the spatial frequency domain.
Product:	[Applications] krita	Reporter:	Robert Scott <bugs>
Component:	General	Assignee:	Krita Bugs <krita-bugs-null>
Status:	CONFIRMED ---
Severity:	wishlist	CC:	halla, miabraha
Priority:	NOR
Version First Reported In:	unspecified
Target Milestone:	---
Platform:	unspecified
OS:	Linux
Latest Commit:		Version Fixed/Implemented In:
Sentry Crash Report:

Description Robert Scott 2007-07-07 17:26:42 UTC

Version:            (using KDE Devel)

Another one of my crazy ideas from Tuesday night...

Make it possible to allow the clone tool to only copy the source's high (or low) frequency data to the destination. This would, for instance, allow different treatment of actual image features and film grain.

The idea came about from my use of the perspective clone tool and its nature of scaling (and hence blurring) the film grain along with the feature, which is usually not what is wanted, and can lead to noticeable discontinuities.

The feature could be further developed to allow band-pass filtering, configurable cutoff frequencies etc.

I can't see the tool being terribly slow, as it can be done with a simple FIR filter and krita must have some pretty good convolution routines already. FFTW might be needed to build the FIR kernel from the brush size & parameters.

But it's the sort of feature where you won't know how useful it actually is till it's implemented. And then it might suck.

Comment 1 Cyrille Berger 2007-07-08 00:10:29 UTC

Just for reference FFT isn't the best choice for Frequency filtering for images, wavelets give much better results (but FFT is still on my todo for KisMathToolBox). As for the speed of the paintop, I am affraid it will be slow, as this can work well only with a big brush. But that is worth playing and trying.

Comment 2 Michael 2015-07-24 01:02:49 UTC

This is a very old request but I think it is an interesting proposal, especially now that Photoshop offers related functionality.

1) Frequency domain operations can produce nice results in certain contexts.  If there is a basis of wavelets that does an even better job that would be quite nifty. https://www.youtube.com/watch?v=ldhG9fmgC7o

2) The complications of creating a localized tool are a bit greater than whole-image operations.  Perhaps a proof of concept through a filter or layer mechanism would be easier than generating a tool.  

3) My opinion is that the most natural domain for these operations is photo retouching and compositing.  Do GIMP or Natron offer frequency domain features like these?

Comment 3 Halla Rempt 2016-01-24 15:14:33 UTC

WISHGROUP: Big Projects