Minimize small variations in color/density?

I take a lot of cloud and sky photos, and often boost contrast. This has an unintended effect. Subtle variations in the blue are exaggerated. (This is not the same as similar post that was more about noise.) None of the methods I have used really worked. Tried creating a blue foreground color, then selecting the existing sky by color and pasting into. Doesn’t work nearly as well as hoped, especially when there is a natural gradient from one section of the sky to another. Comments?

How about some examples?

Here are two…

Blotchy blue sky # 1

Blotchy blue sky # 2

@okieman Could you provide before and afters, and what you did, so that we know exactly what you are talking about?

That image, when enlarged, shows classic signs of JPG compression. JPG isn’t good at recording smooth gradations, and adds noise, especially near sharp edges (ringing). I expect the image looks better with lossless compression.

I would probably try and use a parametric mask, like the ones found in darktable, or a hand drawn mask, to only apply the edits where you really want them.

Since this topic is under GIMP tab, I assume that you use GIMP for your post-processing. I will like to suggest that you use RawTherapee and use the Residual Image tab of the Wavelet module and adjust shadows and highlights. You can adjust the level at which your highlights start and shadows end by adjusting the corresponding sliders. I have had good results with this method.

Later you can also go to the LAB tab in the Exposure module and use LH, CH curves to fine tune your result.

2 Likes

The ‘before’ is somewhere on a storage drive. Here is a zoom in. The dark blue of the sky gets lighter is your eye travels down to the zone between the cloud and the jet trail. I can see a lot of this when I examine the whole image.

blotch

I will try the suggestions from @paperdigits and @shreedhar when I get back! Thank you.

Even though the file is a PNG, that image, when enlarged, shows noise that looks like JPEG artefacts on both sides of the contrail. The PNG file is 8 bits per channel, not the 16 bits per channel I would expect if it came from a raw image.

Is that the problem you are trying to reduce? The first step is to start from a raw image, and ensure JPEG isn’t in the processing chain. Then, if the noise isn’t from JPEG, there are filters in G’MIC that might reduce the noise, but I’m not an expert at G’MIC.

1 Like

@okieman Wavelet decomposition is possible in GIMP as well. See: https://pixls.us/articles/skin-retouching-with-wavelet-decompose/.

Also try using my Darken sky G’MIC filter. It is based on Deep blue sky effect, though it is not exactly the same. (In that thread, I also provided a GIMP tutorial.) I haven’t had the motivation to update the filter to account for over saturation; however, using blend mode L* with the modified layer on top of the original should do the trick.

2 Likes

Part of this problem is that as you get into shadows/darker regions of the image, you actually have less bits available to spread out between shades, and it gets even worse in the blue channel. If it’s posterizing one option to help mask the transitions a little is to add random noise to your image (dithering). I will sometimes use a grain field in Overlay blend mode (there’s a link to a grain image you can use here):

1 Like

Uploaded a PNG that is one segment of original image I needed to extract as a screen capture.
I’m definitely talking about areas of open blue sky, instead of sky that is directly adjacent to clouds etc.

Overall Wow, a lot of ideas to experiment with during the holidays!
Will post results after then.

Or before. So far I’ve only experimented with the @afre Darken Sky GMIC filter. I also created a selection for the blue sky and filled it with solid blue, then set that to about 60% opacity to avoid a fake look, and experimented with luminosity masks. The version on the left. Older one is on the right.

That’s an improvement!