Remove long distance haze

another try…
the other version is way too dark for me …

Remove long distance haze_D200_20070802_2088_01.NEF.xmp (10.0 KB)

4 Likes

Keep up the good work. Love the goatees! :goat:
The challenge is getting rid of the colour cast.

Yes, but the hard part is knowing how to use the raw-black points module.
Rawpedia says nothing about haze: http://rawpedia.rawtherapee.com/Raw_Black_Points
But it seems to work very well in conjunction with the haze removal module.

But … How do you get those 239-530-477 values ​​in this image?

The same here:

Where do the values ​​57-224-219 come from?

Do you have to calculate it by eye?
It must be a very well kept secret :))
Regards!

Changing raw black points alters the white balance as a side effect; however, it is counter-intuitive, so I don’t use this technique very much.

Typically, dehaze algorithms estimate what is called airlight (the particles in the air filter light like a lens) and use that to correct the colour. More advanced algorithms do pre- and post-corrections; and machine learning compares colours against similar subjects.

The RawTherapee poor_man_dehaze branch logs them in console when opening a raw :wink:

1 Like

(Of course, @heckflosse is quite clever. I am waiting for it to reach maturity… :wink:)

Well that is very good and very useful! I’ll wait for it to merge with dev one day … or on 5.9 !!
Thank you.

1 Like

I was wondering this myself, thinking that @heckflosse’s solution was rather nice/clever but maybe not that practical.

Just merged that into my dev branch; That’s a nice one!

At its current info from the terminal state it is somewhat poor-mans indeed, but rather useful in some cases nonetheless! I’m sure you come up with something streamlined in due time :relaxed:

1 Like

dt 3.4.1

D200_20070802_2088.NEF.xmp (6.5 KB)

1 Like

RawTherapee
Thanks for the black points tip… :grinning:

D200_20070802_2088.NEF.jpg.out.pp3 (15.1 KB)

4 Likes

That also works in darktable btw, but there you have to enter the black levels in a different order (at least for Nikon D200 files) which makes it a bit harder…

In darktable for this file you have to use 530, 239, 477, 530

1 Like

@heckflosse

I wasn’t able to use the_poor_man :blush: and i found the black points looking at Art raw histogram. Values of green and blue was a bit different from yours. :slightly_smiling_face:

Here is my take. RT Dev with local adjustments. The key was trying to push the shadows down and pull out some detail, all while trying not to look hyper-realistic by leaving some haze.


D200_20070802_2088-2.jpg.out.pp3 (44.8 KB)

2 Likes

Thanks for sharing! I’m a little bit late to the party, but here is my take nevertheless.


D200_20070802_2088.NEF.xmp (16.2 KB)
and an alternative crop (and a little more pop to the trees)

D200_20070802_2088_01.NEF.xmp (16.3 KB)

DT 3.4.1 I did the dehazing with the dehazing module and local contrast - clarity with multiply blend, then removed the blue color cast with rgb levels

2 Likes

another attempt …

could it be that this picture was photographed through a window pane? or why are there halos at the edges?

translated by g …

Remove long distance haze_D200_20070802_2088_02

Remove long distance haze_D200_20070802_2088_02.NEF.xmp (10.4 KB)

edges halo

Definitly no.

hmmmm …
that would be too even for a ship window.
the clearer the structures and colors are, the more it can be seen around the whole picture. maybe that’s a phenomenon caused by the low contrast of the scene and a filter on the sensor. which is then reinforced and visible through the extremely strong post-processing.

…who knows :upside_down_face:

translated by g …

I wasn’t on a ship. The picture was photographed from here (roughly)

I understood that there is no window :smiley:
(I must learn English…) :expressionless:

automatic translation is not always the right thing to do :upside_down_face:

I recommend DeepL https://www.deepl.com/translator

Better than google at least. :slight_smile: