Jim-Pops wrote:
This was interesting. I opened the image in Photoshop and started worked on it. The picture, as you can see, is very dark. When I brightened it, it bitmapped a bunch, thought I wasn't going to be able to improve it at all. I remembered I have a version of DXO. Tried it with DXO and was surprised to see a major improvement. Saved the improved image then reopened in Photoshop to make the rest of my changes.
Wow. I didn’t even concentrate on the blocking that was occurring in the image. I did know that everything but the sky was in very deep shadow. Other than the sky, most of the image is within the 20 darkest bits out of a possible 256. So the shadows are definitely blocked up, resulting in heavy pixelation as you brighten the image. So I went back and tried to see if the pixelation could be eliminated or significantly reduced.
The first thing I did was to convert the image from 8-bit to 16 bits, but I’m not sure if that actually made a difference.
Then I tried a two step process of blurring the image with a gaussian blur, then re-sharpening (Topaz Lens Blur model). That didn’t work, pixelation was too great for the blur. To eliminate the pixelation, the blur needed was way too much for the sharpening model to have any real effect.
Then I tried PS’s neural filters. Here are the steps I used to get the results below. While not perfect, It did significantly reduce the pixelation and keep much of the image details as shown below (viewed at 200%).
1) brightened the image to the desired level
2) Ran the JPEG Artifacts Removal filter. This took about two minutes to process on my machine (MBP M2 Max, 64GB) so it would take a while on an older machine. This removed some of the blocking but resulted in a slightly blurry image when viewed at 200%.
3) Ran the deJPEG’d image through the Photo Restoration filter, reducing all NR to 0, but setting 'Color Noise Reduction' and 'Half-toning artifacts reduction' to 50. This removed a lot of the blocking that I was trying to get rid of with blurring but without loosing a lot of detail.
4) Final step was to run the image back thru Topaz Sharpening using the lens blur model,