alt Hacker NewsThe line I draw is at semantic editing. Manipulating images based on abstract geometric or statistical features like edges or color histograms is acceptable. Attempting to guess what those features mean is not. Traditional sharpening, color balance correction, focus stacking, lens distortion correction, etc. is fine. Red eye removal, skin tone correction, traditional dodging and burning, replacing a detected moon with a higher resolution photo, etc. is not. Unlabeled semantic editing is dishonest. AI image processing does not distinguish between geometric and semantic features, so it's always unacceptable, even for simple tasks like sharpening.
The one exception I make is cropping, because photographers have always had the ability to choose where to point the camera. Cropping based on meaning is not dishonest because it's inherent in the process of photography.
That doesn't seem like a very sharp line (hah). "Abstract geometrical or statistical features like edges" have semantic meaning. An "edge" is just an area of the image with a high derivative, but as soon as you use that information to process it in some way, you are making some implicit assumption about what this statistical feature means - the boundary of an object, a discontinuity in depth, etc.
There is no dividing line between syntax and semantics - semantics is just syntax scaled. Godel proved it, LLMs exploit it.