New tool "Capture Sharpening"

Yes and no. Let me explain: Current max radius is limited to 1.15 sigma because of performance (1.15 sigma works fine with 7x7 kernels while > 1.15 would require 9x9 kernel).

For the increase it’s easy to allow more more than 50% increase but only up to 1.15 for the reason I mentioned above. For > 1.15 a 9x9 kernel should be used. Of course that’s possible, but not implemented yet.

1 Like

Hi everyone!
I’m Roberto from Argentina and this is my very first time writing here. I’m not entirely sure if this is the right place to post this question… but here it goes.
I’ve been using the latest appimage (manjaro Linux user) for a couple of days, and I just love the capture sharpening tool, but today it just disappeared completely from RT… I deleted the config files, downloaded again… but anything brought it back yet. Has anyone run into the same issue?

Hey @Roberto_Abella, welcome! Can you launch the last workng verwion of the AppImage from the command line and see if it is there?

1 Like

Hi @Mica! I’ve just launched from the terminal and… well… I’m not exactly sure what library I’m looking for in the terminal, but it seems not to be there anything related to capture sharpening as far as I can understand :confused:

Ups… I’m suddenly sooo embarrased… :frowning: apparently this evening when I tried to download the latest build, jus got the wrong one. Just downloaded again a few minutes ago, and there it is…
I’m so ashamed, lol!! I’m so sorry…
Thanks anyway!!! :slight_smile:

No need to be ashamed, we all make mistakes.

1 Like

Except me. :wink::grin:

1 Like

Have you had any luck with this? I am not sure I can do much to help on the programming side, but do have an X-Trans camera and can take any sample images you may need.

Not yet. I’m on holidays atm and will try again when I’m back home at the fast machine…


Enjoy your holidays Ingo!


is capture Sharpening supposed to work on JPGs too?
I open a JPG in RawTherapee-dev-5.7-134-gbbb410f-20191012.AppImage and not only is everything in the RAW section greyed out, but I’m not even able to scroll through it :frowning:

No, else it wouldn’t be in the raw tab.

Yeah makes sense. Thought I’d make sure.

This looks like an excellent new tool. May I ask what the gamma slider is for? I posted a couple of thoughts on RL and gamma here.

One thought is that at low f-numbers the three raw color channels show different sized PSFs because of axial chromatic aberrations even in the best lenses. Typically two of the channels have similarly sized PSFs (G and B, 99% of the time) but one (usually R) is substantially bigger because it is out of focus when the first two are in focus. Also, the OOF channel PSF approaches the shape of a disk rather than a Gaussian.

This suggests adding a per-channel radius option (potentially each also with the option of the PSF being gaussian or a pillbox).

[Ignore the following, probably too complicated, just thinking aloud. Ideally this capture sharpening would be applied pre-demosaic because it interacts with Lateral CA. One could apply Lateral CA corrections to the raw data, deconvolve the individual raw channels each with their own radius, then demosaic and take it from there. Alternatively one could have each PSF shape (gaussian or pill box) affined to an ellipse of varying angle and stretch throughout the image based on the results of the Lateral CA model.]

This is usually only relevant at low f-numbers say below f/ 5.6-8 because after that diffraction becomes the dominant factor in PSF size and shape.


Originally I made the gamma slider to experiment. At a value of 1.0 it means linear processing. In fact, while working on some images in past, I almost always used a value of 1.0 (linear) for my edits.

Hence, let’s vote: Should we keep the gamma slider for 5.8 or remove it for simplicity?

I vote for removing it.


@JackH Welcome to the forum and thanks for your input.

@heckflosse I love simplicity but that is coming from a minimalist. :stuck_out_tongue:

1 Like

Capture sharpening assumes the image is gaussian blurred in luminance channel (by lens diffraction, AA filter, Microlenses). It does not aim to correct for OOF or channel dependant PSF.

Of course, that’s not perfect, but it seems to be a first step in the right direction (at least I think so and the amount of detail I can get out of my old D700, which has an AA filter, files, does not prove me wrong).

Applying capture sharpening on RGB instead of L(Y) would mean increasing the processing time a lot (almost 3x)


1 Like

Some thoughts about capture sharpening:

  1. If someone wants to port it to DT, it’s an almost perfect candidate for OpenCL usage as the main step is to transfer the data to GPU memory, then do the iterations, then transfer back to main memory if needed.

  2. in RT 5.7 dev it’s still wip, means that behaviour can change for 5.8

  3. The ability to process tiles with different settings (Corner radius boost) may me explored a bit more…

1 Like

I stuck a gamma parameter in my filmic tool because that’s how Hagle presented it in his blog, needed to compare results to Reinhard. I never got around to messing with that, so it’ll probably disappear in rawproc 1.0. I’ll sill have to parse it, got about a year’s worth of images with that in their toolchains…

I’m a fan of “one tool, one purpose”, so I’d vote to remove it, if the vote of someone who doesn’t regularly use RT counts… :smile:

Sure it counts!

1 Like