Fixing severely under AND overexposed images

[camner]

Operating System

1. macOS 15 Sequoia

I have always struggled with editing images that are severely underexposed, and struggle even more when the same image is both under and overexposed.

This image is a scan of a slide taken with my first "real" camera purchased in 1980. It was one of the first shots I took with the camera, while on a short trip to Williamsburg. The main purpose of the trip was to have an opportunity to practice with the new camera before taking it to Europe later in the summer.

At the time, I really didn't understand about various metering modes, so left the camera in the default "Evaluative" metering mode.

Since there was light streaming in through the window, I managed to get a shot that was both overexposed and underexposed.

To tone down the brightness of the window, I selected the window as an object and reduced both highlights and exposure. That helped a lot, and I think if I blur the background later, it will look even more natural.

I selected the woman as an object and increased both shadows and exposure. The problem I have noticed when working on significantly underexposed portions of an image is that increasing shadows and exposure results in "fog" (or "haze" if you prefer). Increasing Dehaze helps quite a bit, but still results in an image that just doesn't look very natural.

I know there is a limit to what can be done when the starting point is particularly poor, but I'm wondering if there are some more subtle tricks that I don't know that might do a better job of salvaging the image? (This particular image is not important to me, but there are other images that show the same issue that do matter to me.)

Original tiff scan of slide:


After my adjustments:

 

[Johan Elzenga]

I would not try to fix the overexposed window. Just leave it like that and concentrate on getting more detail in the shadows.

 

[camner]

I would not try to fix the overexposed window. Just leave it like that and concentrate on getting more detail in the shadows.

Thanks!

The reason I wanted to do something with the overexposed window is that one's eye naturally tends to see brighter parts of an image first (isn't that why vignetting is used, to draw the eye to the subject rather than to another part of the image that happens to be brighter?).

Any thoughts about how to avoid (or deal with) the haze that gets created when one increases shadows and exposure in a severely underexposed area (beyond using Dehaze, which helps but doesn't really resolve the unnatural look created by pushing shadows and exposure)?

 

[Johan Elzenga]

You could consider changing the image to B&W. I’m on my iPad right now, so I have limited options to try something myself, but I think that is already an improvement. The light from the window is only natural. The viewer will expect it to be the brightest part of the image.

 

[camner]

Thanks for the B&W idea…I don’t usually work in B&W, so I am ignorant about when to do so results in a better image. I will give it a try.

 

[Conrad Chavez]

The potential for improvement can have a lot to do with how it was digitized, so some important questions:
What kind of film was it, transparency or negative?
How was this digitized? By scanning a print, or with a film scanner, or digitized with a camera?

There will be the most chance of recovering image quality if it was negative film scanned with a film scanner.
Transparency film has a narrower dynamic range than negative film, so getting much more out of the shadows is a bigger challenge with transparency film.
A film scanner with a higher dynamic range sensor might be able to get a little more out of it, but probably won’t work miracles. Some film scanners have an “analog gain” feature that can sometimes help with shadow detail.

I don’t think it’s necessary to go full B&W, but desaturating might reduce odd color shifts you can’t easily resolve. I cut Vibrance by -20.

I wouldn’t say my attempt is perfect, but there is highlight detail to recover in the window. Of course, I used a number of masked adjustments, especially a radial gradient to specifically address the window light. Also, some of my color balancing was done with RGB curves because on scans, those can often be more useful for color correction than the Temp/Tint options.

The demo below starts with a quick Before/After before showing edits. The image quality of the demo is slightly reduced from the original screen recording because of the limited color palette in a GIF animation.

 

[camner]

Thanks for taking the time to do this. The gift showing them before and after is particularly effective!

You've given me some ideas about how to approach correcting this kind of image issue as well as introduced me to some tools. I haven't been using (for example, RGB curves).

To answer your questions, the original is a Kodak transparency. The scanning was done by a outside service using a Nikon Coolscan scanner. I don't know what settings were used.

Thanks again.

 

[Conrad Chavez]

To answer your questions, the original is a Kodak transparency. The scanning was done by a outside service using a Nikon Coolscan scanner. I don't know what settings were used.

That’s good, because if they used a Nikon Coolscan that should be at the higher end of the potential scan quality you would get out of transparency film, even at default settings. So even if the scan file doesn’t look great, there probably isn’t very much that could be done to get a better scan.

 

[Swingman]

My first SLR was Pentax KM in 1975. I was also new to photography and those cameras had only manual exposure and manual focus. I also used film initially and then switched to slides. I tried to copy them using Samsung S24 and found specially the slides had very small headroom to make meaningful corrections. The initial results were poor. They also had colour cast (ageing) and struggled to rectify it. I only converted about 100 images and stopped it for a while until the later in the year. There were few suggestions in this forum and will try them and see how much improvement it makes.

 

[camner]

@Conrad Chavez
What do you use to make a gif of your screen?

 

[RoyReed]

Could you get the scan done again, this time exposing once for the room and once for the window, and then doing an HDR merge?

 

[Conrad Chavez]

What do you use to make a gif of your screen?

I record using the built-in screen recording feature in macOS (Command-Shift-5, similar to the Windows-key-Shift-S shortcut in Windows 11). Then I run the video file through Gifski (free) because it does an unusually good job of optimizing and dithering video colors down to the 256 colors that GIF is limited to.

For posting, it's important to keep the file size down. For example, this forum limits an attachment to 2MB. If you want a GIF animation to fit within that limit, you have to intentionally do things that allow a GIF to compress well, like:

• Minimize the dimensions…make the capture area no larger than necessary.
• Minimize the frame rate. I had to reduce the frame rate of the demo above to 4 fps.
• Use as few colors as possible.
• Use as short a duration as you can.
• Try to change fewer pixels from frame to frame; frequent whole-frame changes will explode the file size.

Could you get the scan done again, this time exposing once for the room and once for the window, and then doing an HDR merge?

It's a good idea, the Nikon Coolscan supports Multi-Exposure which is a form of bracketing the scan exposures. How much more they’ll get will depend on which transparency film it was, how many stops of dynamic range that specific film type could actually capture.

I have some similar-looking exposures from my early cameras and might see if it makes a difference with my scanner.