Capture sharpening v contrast/details v wavelets/contrast

Software RawTherapee
1

RT 5.8 appimage

I’m lost, mostly because I don’t understand the differences between:

  • Capture sharpening
  • Contrast/details
  • Wavelets/contrast

I have looked in rawpedia, but I don’t understand which I would use.

Any guidelines on when to use them?

Thanks

1 Like
2

Capture sharpening is Richardson-Lucy deconvolution at the earliest possible step in the pipeline. Its basic premise is to combat lens blur without the artifacts if you use RL sharpening later in the pipeline. This is the only one I use. In fact, it is in my default workflow.

The other two I only know conceptually because I don’t use them. Wavelets might be what I would prioritize next because it separates the image into various detail scales, so you could focus on enhancing certain details and contrasts.

It isn’t so much a versus thing; it is more about what you want to do with your image.

2 Likes
3

By Contrast/details I understand that you mean Contrast by Detail Levels.

Following the explanation given by @afre, I wish to further differentiate between them:

  • capture sharpening: it’s always a must use tool
    • if you don’t resize your image upon final exporting, then this tool is about all you will need to sharpen your image (always depending on the camera/lens combination inherent sharpness, and on your skills while shooting). Perhaps a bit of extra standard sharpening will be needed, but not much
    • if you resize the output image, then you have to look at this tool as a helper for other tools, so they work much better, but most probably you will always have to post-resize sharpen the output image
  • contrast by detail levels (CbDL) is a tool that uses wavelets (D2 wavelets) and only allows you to tweak the finest details of the image. Sometimes this is enough, others don’t.
  • contrast module within wavelet levels tool: this tool is more comprehensive than CbDL, as it allows to choose between D4, D6, D10 and D14 wavelets, you can tweak up to 10 levels of detail, plus a residual image (what remains after extracting all the details), and has many more modules to work with each level and combined with the contrast module

As a general rule, you may wish to always use capture sharpening (fine tuned for each image) and wavelet levels contrast (alone or in combination with some other modules present there).

The problem is that currently the documentation for wavelet levels is not up to date in English, so perhaps you may get decent enough results with CbDL until the documentation is finally updated.

On the other hand, you can sometimes combine capture sharpening with other tools (Tone Mapping, Local Contrast, …) and get a perceived sharpness increase that feels enough for you.

1 Like
4

Normally, I use (in this order):

  1. capture sharpening with usually a much higher contrast threshold (according to the preview) and also usually a negative corner radius boost(I’m usually no interested in sharpening the corners) +
  2. sharpening with a high contrast threshold according to preview 1:1 +
  3. micro-contrast: same contrast threshold as sharpening +
  4. contrast by detail levels (around 1.9) +
  5. post-resize sharpening same contrast threshold as sharpening and amount around +33 or +44

But my sharpening is mostly for dogs and buildings: “crisp is good” :slight_smile:

Here’s an example: The Dogs of Vancouver – Happy Dog: Vancouver, BC

I’ve tried quite a few times to use the wavelets, but since I don’t really understand it and have not found good videos or simple explanations. Usually, I make it worst and/or unpleasant.

Thanks

5

As I see it, you sharpen mostly the most pronounced edges.

If you downsize an image, the enhancement done with microcontrast is basically lost. If you need that extra crispness, you may wish to apply it after downsizing the image.

The wavelet levels documentation will find its way into the website. It’s just a matter of time

6

When applying multiple modules, observe how the sharpening stacks and interacts. Artifacts will become progressively worse. Not saying your workflow isn’t good for your purposes but I am a little concerned about the use of every single tool in the toolbox to change a light bulb so to speak. Maybe try using less tools but at more aggressive settings, and without duplicating a similar effect, then compare and contrast. E.g., I would push capture sharpening a bit more.

1 Like
7

Here’s what I used for this image:

Black & White
Crop
Directional Pyramid Equalizer
EPD
Exposure
Film Simulation
HLRecovery
Impulse Denoising
PostDemosaicSharpening
PostResizeSharpening
RAW
RAW Bayer
Resize
SharpenEdge
SharpenMicro
Sharpening
White Balance

https://www.sritch.com/dogs/dodoite-happy-02-vancouver-bc.html

How to push the capture sharpness? Right now I’m using:

[SharpenMicro]
Enabled=true
Matrix=false
Strength=20
Contrast=35
Uniformity=5

Thanks

8

I am not feeling well ATM. Take a look at Pipeline sequence in RT. This would be my approach to sharpening.

1 Favour steps that are closer to the start of the pipeline. The earlier the sharpening the more liberal you can set the parameters. Double check various image features for artifacts to make sure your values aren’t too high.
2 Sharpen before or after steps that may soften the image but only enough to recover from the softening.
3 Once you have done steps 1 and 2, then and only then adjust for (artistic) taste.
4 Compare, contrast and have fun!

9

Sorry, but I don’t understand what you mean. Can you explain?

English is not my first, nor second language.

Thanks

10

I think there are a lot of forums and discussions about the need or not need of what is called pre-sharpening.

What I think @afre wants to tell is that the closer to the start of the pipeline you apply an edit that enhances the image (that looks «sharper» to your eye), the fewer the artifacts you will introduce (always taking care not adding too strong settings in the tools used).

Then there are some editings that soften the image (a clear example is downscaling/reducing the image), and you may wish to sharpen the image again (a bit) with a tool that is applied later in the pipeline, so you can recover from the softening.

Ideally that would give you a neutral image, without too much sharpening, and without artifacts introduced by sharpening. Then, you could add as much enhancement as you like.

1 Like