Happy new year, hive mind! I have a question.
I'm a fan of George Inness and am trying to see what this painting would have looked like before the ravages of time made it so gloomy 🙂 
My impulse was to go to adjust brightness and contrast, levels, and hue/saturation — but the results were disappointing—more legible but it seems really washed out. (See my file below with non destructive adjustment layers)
Any thoughts on a systematic way to do corrections like this?
Thanks in advance.
Explore related tutorials & articles
11
Replies
The problem with the original is the blue channel. It's clipped to the point where there's no information left:
The lack of blue channel information is what gives the heavy yellow cast. To fix it here, I had to borrow some from the other channels, with a Channel Mixer layer. Use Blend If to constrain it to the low end of the underlying blue channel. Note - the proper fix would be to go back to the raw file and retrieve as much blue channel information as possible there.
Then it's a simple matter of using a Levels layer to lift the blue channel and (a little less) the green channel. Just drag the high endpoint.
Finally Selective Color as a finishing touch, just to reduce yellow a tiny bit more.
Above is my result.
This was done using a very clever and super fast technique that I learned from some Adobe video editing teachers, where if there is a single dominant unwanted color cast in an image, it can be cancelled out in a few seconds by adding a solid color layer set to the unwanted color, and applying the Divide blending mode to it.
These are the steps, also used in the Photoshop demo below:
1. Create a new Solid Color fill layer.
2. Hide it.
3. In the Layers panel, double-click the Solid Color fill layer, and click its eyedropper on a part of the image that has the unwanted color cast. By the way, this is the reason for step 2, because if the default solid black color is not hidden, you can’t see what to click in the image.
4. Unhide the Solid Color fill layer, now that you’ve set the color.
5. Set the layer blending mode of the Solid Color fill layer to Divide.
6. (not shown) If you don’t like the results, you can double-click the Solid Color fill layer to adjust the sampled color.
The image quality is not good in the demo below because the Adobe forum software is not kind to colors in uploaded animated GIF; my original of that GIF looks fine.
This was done using a very clever and super fast technique that I learned from some Adobe video editing teachers, where if there is a single dominant unwanted color cast in an image, it can be cancelled out in a few seconds by adding a solid color layer set to the unwanted color, and applying the Divide blending mode to it.
By @Conrad_C
Thank you for sharing Conrad!
This reminded me a bit of the old colour cast removal technique of creating a solid fill using the inverted colour of the unwanted colour cast colour, then blending in color blend mode at around 20-30% opacity.
This was popular before Adobe added the "Photo Filter" adjustment layer such as "Cooling Filter (82)" or using a custom colour value.
The results of the divide method are very different though. This is a good thing if using the "variants" approach to combine different quick corrections into one.
Firstly, this is very much subjective. There are of course general guides that one can use for sky, clouds, grass or foliage – this is very much a case of deciding on what your "vision" should be for the colour and contrast. I am not familiar with this artists works, so there may be a common theme that you can use as a guide.
One can use many different techniques, which is an excellent thing (more on this shortly).
Although we can discuss some of these individual techniques as @D Fosse and @Conrad_C have done, I'd like to start by moving beyond individual correction methods.
Traditionally, one puts in as much time as commercially or practically possible in a hobby or labour of love situation when working on an image. Attempting to make a "perfect/great/good/acceptable" overall image in the allotted time.
When Dan Margulis introduced the "variants" concept, it changed the way that I and many others approached image editing. This workflow proposed making multiple versions, using different "fast" corrections – as opposed to spending a long time on a single series of edits.
Let's say that you only have 20 minutes of "budgeted time" for a correction of a particular image. The traditional approach is to spend 20 minutes on that image, making the best that you can. The "variants" approach could be to edit 2-4 different versions of the same image that take < 5 minutes each. Some attempts may be better in some areas than others. They may also be similar but subtly different. One can then intelligently combine the different images taking the best areas of each to create a composite – or one can simply average all of the variants into a new combined version. This average version could then be the final image or the starting point for further "traditional" fine-tuning.
Here are 7 quick variants, in no particular order or preference:
This is what Dan termed the "Par version" – or a simple average of the variants:
And here is the original and the "par" version (average of all 7 variants):
You may or may not like the final "par" version, or perhaps the "par" would be better if averaged from only 2 or 3 of the variants (culling unsuitable variants). Or perhaps rather than using the "stupid blend" (the mean/average of all combined images), it would be better to layer in different opacity or blend modes and or mask and combine different areas of the variants to form a composite image.
Remember, the combined result can just be the starting point for further editing. In this example, in the available 20 minutes, one may spend 15 minutes making variants, then combining the "winners" and then spend the last 5 minutes fine-tuning as one may traditionally do from the original unedited image.
Although Dan goes into this in more detail in his books, videos and PPW PDF documentation, some freely available links on this topic include:
http://www.moderncolorworkflow.com/blog/the-mit-5k-dataset-3-effective-averaging-close-up
http://www.moderncolorworkflow.com/blog/the-mit-5k-dataset-4-more-on-averaging
http://www.moderncolorworkflow.com/blog/ppw-blending-strategy-in-a-flower-image
https://youtu.be/MdQb-iZYDro?si=c_XeYlcLWeaFUk3N
Wow! Thank you for the detailed response. I haven't experimented much with the variants function but I'll give it more time in the future 🙂
By @Noyster2
There isn't a specific "variants function" in Photoshop, what is meant is that you make different corrections, then combine them.
Take the following 4 "variant" examples:
1) Using the Camera Raw Filter to "fine tune" the image using the various sliders (working on a smart object copy)
2) Using the method that @Conrad_C proposed
3) Using the method that @D Fosse mentioned
4) Using some other series of edits to create another version.
So now you have 4 different corrected images, each may have pros/cons... But by combining them in different ways – "the whole is greater than the sum of its parts".
Opening image with Camera Raw and Auto correction doesn't work so bad…
If not I would use curvbes and set up of white and black point…
AdChoices