I just returned from a camping trip next to a lake and took some birding photos. Some of them were a bit OOF and I decided to see what Sharpen could do. On the most part I was amazed at how well the 3 modes did with some of the photos, but then I came across this:
Here is part of an image I caught. You can see only the part including the head because that is what I wanted to sharpen. This is the image in Sharpen after adjustment but before saving it. For what it is worth the image looks a lot better than the reduced version shows.
However here is the image after save:
<img
I have no idea where all of that distortion came from, but I tried it using both Focus and Sharpen and both show the exact same artifact. It can not be because there is too much adjustment as I only used 30% strength and 20% noise reduction.
I have been more and more impressed with Sharpen as I worked through my images, but this was a big disappointment.
I’ve seen this distortion happen too on skin tones. Usually it is apparent in the preview and when this is the case I cancel AI and go to traditional sharpening methods. Artificial Intelligence is as the name suggests, artificial, and often it shows. As the science progresses you’d expect these anomolies to get less and less but it’s still in its infancy (though I remember doing a pn Open University module on AI back in the 90s). Hopefully these anomolies will lessen as new algorithms are developed.
I have two workflow tools that I use for editing. I was using the “lighter” one, Dxo’s Photo Lab, when I encountered this issue. I had the software process the raw image as a 16 bit uncompressed tiff to Sharpen, adjusted it in Sharpen (where I did not see the anomaly) and save it back (where I did see it).
This morning, out of curiosity, I used the other tool, CaptureOne, and repeated the process. That is, I sent an uncompressed tiff to Sharpen, adjusted it and saved the result back. I did NOT see the same issue with CaptureOne although I do not know that the source of the tif is the reason. Both images were cropped and perhaps the second crop was enough different from the first to prevent the issue from showing up.
Thought people might want to know that perhaps different source software may result in different results.
I’m wondering too that if images passed to Topaz AI and edited in other apps have an adverse effect on the AI algorithms. The evidence suggests that it might.
Yes. I think the quality of the tiff that is sent to Sharpen may have a significant impact on the ability of the app to process properly. So far I have only tried Dxo’s Photo Lab, CaptureOne and Photoshop (as a plugin). I think I need to do some additional testing.
<img