RT vs DT - sharpening - 1001

Oh please don’t get mean to me, there are so many comparisons to this, a newcomer will probably be killed by it.

The simple sharpen module in DT (radius, amount, treshold) works so incredibly well for all my pictures. In RT I can move days with modules, in DT one is enough.

How can I map the simple sharpening of DT in RT, even if it has always been discussed that RT is much sharper?

Hi @bediicco,

It’s funny, but in my case, I feel that I can get
a nice development faster using RT than what
I can get in dt :-)))))

Have you read the wealth of valuable info in RawPedia,
and watched all tutorials on YouTube?

Have fun!
Claes in Lund, Sweden

1 Like

I think modules named “sharpen” in both apps are disappointing. RT’s capture sharpening is nice and so is the RL resize sharpening.

In dt, its two instances of Diffuse or Sharpen, one for AA filter sharpen and one for Local Contrast.

1 Like

If it works for you, that’s ok too.
RT has a few algorithms available at various points in the processing pipeline.
It essentially comes down to the algorithm and order of operations.
The basic methods we have are Unsharpmask and Deconvolution.
They react differently to noise, and this is furthermore sensitive to the debayering method used. If I have something I can sharp, I will try deconvo first, adjusting the thresholds as required by the noise floor.
Sharpening in RT can be applied to the RAW input, the debayered image, and the output image.

Hey @Claes

thank you, no, i have not the time to read and watch all the day long.

Thats why i am talking about :wink:

Cheers

i have not the time to read and watch all the day long.

Simply put: Oh? Then you will never become sufficiently skilled…

2 Likes

DT can often look good with sharpen at first but you need to look for halos and check at 100 for the real impact as that is what you will see in the export… noise and sharpening are shown at zoom less than 100 in DT but often not that representative of the final output. Diffuse and sharpen is cpu hungry but can yeild pretty amazing results… RT only shows these at 100 percent so sharpening might appear to do nothing unless you zoom in

1 Like

I have such a good workflow in RT, it is so accurate and extremely fast, but the sharpen is one of the things that are extremely “fast” (easy) for me using DT.

In most cases in RT i have usage of unsharpmask which gives good results, but deconvolution does mostly not work for me. It is morely like a very very small camouflage effect…

That’s the old RTFM bias… play_raw will get you there faster.

1 Like

We would have to see screen shots, RAWs and .pp3’s to understand what you are stating about the performance of the shapening tools.

I would like to suggest that in Darktable the difuse or sharpen module (D&S) is incredible for sharpening rather than the old fashion method of unsharp masks which the sharpening module in DT and most programs is based. In the D&S module I initially apply demosaic sharpening (AA filter) pre-set. This is sufficient for many images including portraits. However, if I want or need more sharpening I then apply a new instance of D&S using one of the lens deblur options. The soft one is sufficient for most images but sometimes I use the stronger ones of medium or hard. This is my new sharpening method and I just love the results.

BTW, there is a details threshold slider in the masking option that allows sharpening to be restricted to edges and details rather than the whole image. This minimises noise problems.

2 Likes

@Terry

Thank you i will follow your guidiance the next days and try for me

EDIT: Something may have gone wrong in the process here, I’ll have to check that again first

Note for me: X3 vs X4

Try the no AA filter preset… I often dial it up to 2 or 3 iterations…more and it can become too much… but maybe you would have a different tolerance… maybe keep going and then back it down…for me 2 or 3 iterations is a sweet spot. I also used to use the dehaze preset in D&S and found it to add an very nice degree of sharpening…

2 Likes

No matter what materials one uses to learn, it takes time. ‘i have not the time to read and watch all the day long.’ - going through the PlayRaw sidecars, watching videos and reading all take time.

2 Likes

Then your experience is very different from the typical — sharpen should not be used (artifacts), Ansel went so far as to deprecate it.

As others have suggested, diffuse and sharpen is the preferred solution.

For years I taught digital imaging in programs like Photoshop and GIMP and the unsharp mask was the gold standard. However, people like a sledge hammer and over use the unsharp mask. I do feel the tools created by AP are just brilliant and make the unsharp mask look old technology which it is, but I hate to see modules deprecated unless they cause maintenance issues. The Diffuse and sharpen module is certainly resource hungry, but really worth it.

All code causes maintenance issues in the long run. Even if its functionality remains the same forever, API changes will require refactoring. Deprecating and then dropping modules is the only way out.

FWIW, I think that not deprecating modules which have a better alternative is confusing to the user. Hence the “which one should I use” questions, followed by the “why is it then still around?”. Both of which are fair questions, with deprecation as the answer.

Regarding speed: you can use a single iteration. For some adjustments a “fast” preset exists. It will not be pretty for major adjustments, but I guess it will be superior to plain vanilla sharpen (counterexamples welcome).

I defer to the generous developers when it comes to this issue. I have no right to complain unless I am willing to learn coding and acept the onerous task of maintenance. BTW, my windowws laptop died last week and if I can not repair it I am planning on getting a new laptop with 6GB graphics card just for the diffuse and sharpen module which I love, In the meantime I am using a linux desktop. I love how linux makes old computers new again.

1 Like

Same here, and it should be noted that project governance is especially hard in FOSS. The decision to pull the plug on a feature usually does not happen easily: someone might be using it, it represents someone else’s effort, etc. While the status quo is comparatively easy to stick to.

And to be fair, DT devs removed the auto-apply sharpening in 4.2, so I hope sharpening will be deprecated at some point. But I could not find an open issue about it.