Here is a bit trickier care shown step-by-step. Please refer to my FOSS HDR panorama tutorial to understand how I work in basic cases.
In this case the challenge was to high contrast of the scene: the lighting I wanted was just daylight shining through a single window, shining to almost black room. And the angle I wanted pointed straight towards the window. As you can see below, overexposing did not help; on my compact camera the scene was flooded by overexposed areas long before I could capture details in the shadows:
I saw this at the time of shooting already so I shot another panoramic HDR set, this time with the light on. I originally shot 9 images for each 6 HDR frames (3 for each two panoramas) at 1 EV spacing, giving me source images from 1/15s 15s but because of the strong bleeding I could not use the 15s one. Usefulness of even the 8s one is up to a debate - it brings a lot of noise and very little information to the stack. Here you can see both versions with HDR and panoramic assembly done:
I then ran the images through my regular set of tone mapping algorithms: reinhard05, mantiuk and mantiuk w/ contrast equalization. Normally I stack those three with the original HDR version and pick something out of each, but this time I used only the mantiuk+CE images because they matched the feeling I wanted to get almost perfectly. Below you can see the Mantiuk+CE versions:
Now the problem was.. I had two images, one with the feeling I wanted, another with details captured. But, having foreseen this, I had processed them with Hugin in a same project, so I could stack them and they would match features almost perfectly.
So, I started by putting the lit version at the bottom, to have the all the details captured. I reduced saturation on yellow tones to make it match more the lighting I wanted (Hue/Saturation -> Yellows in Photoshop):
Next I dropped the non-lit version on top of it, and picked “Color” as the type of the mixing. As you can see below, it’s already quite good:
Next I wanted to make the lighting more dramatic, like it would have been if I had been able to capture the version I wanted properly. I played around a bit and found that putting another copy of the non-lit version to the top with some darkening filter worked well. I ended up using “Hard Light”.
The problem here was that it also greatly enhanced the noise in the areas where the non-it scene had most of it. I solved it by creating yet another copy of the non-lit layer and added enough median filter to it to get rid of almost all the noise. Now the result was a bit oil-paint like but noise was not a concern. Because the median filter was only applied to “Hard Light” layer, there was still details in all the areas from the lower layers. I already further processed a ready-to-be-published version from this but then I started to have second thoughts. The solution was to use both the original non-lit image and the median-filtered one as “Hard Light” layer and to use masks to pick which one of them to use. Below you can see median-filtered (on the left) and the original versions being used as “Hard Light” layer:
And here’s how it looks when I have painted a mask that picks just right version of either of those two versions around the picture:
Next I wanted to work with color correction so I duplicated the image, flattened layers and saved with new name. I started by removing some green fringe. I picked the approximate area as selection and used “Hue/Saturation” again, now on “Greens” and reducing saturation to zero. Here are before and after crops:
Now only thing I still needed was to crop the image down and give it some more contrast. I used Photoshop “Shadow/Highlights” dialog that hides very useful “Midtone contrast” tool. I very rarely use the shadow/highlight tools, but midtone constrast I use on almost every image. Here’s how the final image looks like, cropped and midtone contrast enhanced: