PROCESS
>hot pixel filter
>changed the demoz algorithm to IGV to reduce maze-like noise patterning
>highlight recovery
Used the color propagation mode of highlight reconstruction to keep as much color detail as possible, then used a mix of exposure compensation and highlight compression to remove the clipped areas.
>white balance
Used the pipette to pick a spot from the whitest-looking window. I used a spot around the edges where the light was less intense, more "neutral."
>noise reduction
Default settings.
>tone curves
I couldn't get much out of the shadows without going out-of-gamut but I was able to lift the midtones and highlights (see: ibb(d o t)co/sv1cdxJJ). I used the saturation and value blending mode on the first curve again to increase contrast without adjusting saturation too much, then I used a weighted standard curve on the second curve because it *does* contrast and saturation together, like the standard curve, but it's more subtle. Read the exposure page on Rawpedia for more info. I had to turn saturation down by -39 anyway. These high ISO images seem to bake hard.
>color toning
Used the hue-saturation vectorscope as a guide to decide what colors to start with but after that I kept my eyes on the photo and adjusted the position of the points of the two curves and the curve "weight" (how much it leant to one side or the other) until it seemed "good" (see: ibb(d o t)co/zhWR3bFy). The out-of-gamut color indicator keeps me from going too far, I think. I would probably get more high-key results just following my eye.
>local contrast
Used a smidge again. Again, the out-of-gamut color indicator reins me in on making adjustments. As I left it, I could barely see any difference.
>LAB adjustments
Used the L* curve to lift the lower midtones (actually the spread of light around the windows).
>impulse noise reduction
Removed salt-and-pepper noise.
NB: you can only see noise reduction adjustments at 1:1 view or more.