I’m new to RawTherapee and would like to know if I am understanding the noise reduction correctly.
I followed the Rawpedia documentation by using the Luminance and Luminance Detail sliders: Noise Reduction - RawPedia.
However the result is visibly not as good as what I achieve with Lightroom (using the exact same 2 sliders).
It is especially the case for high iso (3200 ISO) and on people skin/face.
I know that there are a lot more options I can play with in Rawtherapee (wavelets for example) but I was amazed by what I can achieve so simply with LightRoom and wanted to know if I could do the same with Rawtherapee.
I think I have found why LightRoom seems to be so much more efficient.
The starting points (before any noise reduction) of LightRoom and RawTherapee are really not the same.
No, I don’t think so. LR does some things behind the scenes even for ‘starting point’. You never see the ‘neutral’ (whatever ‘neutral’ means) raw file with current LR.
I think you can see the same image like RT if you disable the module ‘detail’ in LR, or put the sliders ‘luminance’ and ‘color’ in ‘noise reduction’ to 0. The first image has chroma noise removed.
Isn’t getting an equivalent final result better than getting an equivalent starting point?
Btw: you should at least enable ‘Impulse Noise Reduction’ for your image. That may also allow you to reduce the non -impulse noise reduction a bit
You are right, however I haven’t managed to get a good result with RawTherapee. So I am trying to understand why.
I’ve heard that the noise reduction is really good, so I was surprised that I couldn’t get close to LightRoom (even though I spent only 1-2 minutes on the LightRoom noise reduction). I though that perhaps I am doing something wrong.
Unfortunately, Impulse noise reduction has no effect on the image I got.
Lightroom baked “tweakings” have always been shameful no matter what camera, no matter what version. LR user since beta 1, almost a decade. But hey sometimes one doesn’t have toeven proccess the file, already graciously developed by the dark force, ejem adobe inc XD
///// If you’re on a smart car or browser gets sad, here’s a DL link ~43MB
To the admin. If file too big, I can make 1:1 crops, no prob, despite the image’s of no interest, I thought may reflect better the overall impact {come on here doggy}
Everyone wait.
I think I’ll make an example with a real human face to show the obvious difference.
I don’t understand why, but it’s particularly noticeable on human faces.
The first thing I noticed that is your RawTherapee’s output is much brighter than Lightroom especially in darker areas, which will make the noise look a bit more apparent. So I tried to bring the overall lightness in line using some curves, it is not the same but should be closer.
Now onto the actual noise reduction.
I always deal with the chromatic noise first, and a good starting point is using the Automatic Global method with Quality set to High. This should be the two click solution to deal with 95% of the color noises on almost any RAW images, and it is also very aggressive, so I lowered the Chrominance Curve quite a bit to avoid destroying the colors. You will probably notice some random 1~2 pixel-wide colorful blobs left in the image if you pixel peep; this is eliminated with a Chroma Only 5x5 3-iteration median filter.
With all the color noise gone, we can deal with the difficult part: luminance noise. The trick here is setting the Luminance slider to 100, and then raising the Luminance Detail slider to bring back some details. I find this much more effective at combating high ISO images and now you only end up with one slider to control the noise/detail ratio. In my example I set this slider to 55.
Next, I turn on Impulse Noise Reduction with threshold 75 to further remove noise in the darker areas, and enabled Microcontrast to bring out the really fine details that were lost from all those NR. Finally, I used Edge Sharpness in the Wavelet tool to sharpen the edges, along with edge detection to avoid sharpening the noise.
I think the result should be fairly close to Lightroom now. It is not a two-slider solution like LR and takes some experiment to get there, but you can save all these in a preset and then just adjust the luminance detail slider for different images.
Thanks for the attempt !
It is really more obvious on people skin, but if we zoom a little, we can compare the result between LightRoom and your RawTherapee processing (left if LT, right is RT):
The result with LightRoom seems to be more pleasant to the eyes.
Again, I don’t want to appear as a pixel peeper/snob zooming like this, but the difference is really visible without zooming too much with people face.
I don’t really mind working a bit more with RawTherapee to achieve a similar result with LightRoom. But so far even after having spent hours, I couldn’t achieve as well as 5 minutes with LightRoom (as a total noob, I only installed LT to have a point of comparison, but never actually used it before).
I must be doing something wrong, because nobody else seems to have problem with RT noise reduction.
[quote=“Jessie, post:16, topic:2887, full:true”]
The result with LightRoom seems to be more pleasant to the eyes.
Again, I don’t want to appear as a pixel peeper/snob zooming like this, but the difference is really visible without zooming too much with people face.[/quote]
To be honest, I prefer the RT result, although there is still some more grain. The LR version looks flat (washed out) to me regarding color and contrast.
If you want to compare noise reduction and make statements that the output from one program looks more flat than another, then you should first make both programs handle the image as similarly as possible. Specifically you should set LR to neutral, then set exposure compensation in RawTherapee to match LR, and then use a tone curve to finalize look. LR’s neutral uses a s-shaped tone curve, RT’s neutral is really neutral. The PP3 I downloaded from the first post has exposure compensation set to +1EV, flat tone curves and it does not use a DCP input profile with tone curve - that alone means you are not comparing apples to apples. Spot-check pixel values to make sure they match between both programs. Once you do that, then you can compare noise reduction and see which version is really more flat.
… yes, and once the two images are similar, you can use “ImageJ” and start an analysis of the mean, standard deviation and Kurtosis of an identical line or area on both images.
I’ve looked at your pp3 and noticed that you used the s-shaped tone curve. However you haven’t used a DCP input profile with tone curve. Did I misunderstand your explanation ?
arg mas non!!!
