>>3734813>you will never, ever find a hard one pixel line that follows an exact horizontal or vertical in a photo.Mostly true because we're all using bayer filter digital shit so even when the lines align well geometrically we still see a fuzzy 2x2 pixels wide, at minimum, blurred line on the pixel scale.
Most cameras have an anti-aliasing filter too so images 1px sharp never even make it to the sensor either so there's softening going on throughout the whole system, and good software rotation downsides will be barely visible. An aligned horizon is worth it.
>>3734779Your examples are artificial.
Yes, rotating reduces sharpness, but in practice using decent software the clarity lost is irrelevant because your source isn't perfectly sharp to begin with.
If you're scaling down your 24MP shot to 1500x1000 then rotating you're an idiot. Rotating a full res 6000x4000 image before scaling down however is fine. Or just a 24MP rotated image displayed at 1:1, that's fine too.
Your examples are judging pixel perfect images not resampling well. This is an almost ideal example to showcase the underlying issues but fails to take into account that digital pixel art is essentially a perfect image displayed back perfectly and is completely impossible to match with imaging devices.
Look at
https://en.wikipedia.org/wiki/Optical_transfer_functionIn reality...
No lens is perfect.
No sensor is perfect.
Even a theoretical perfect sensor is gonna be gimped by a color filter 99 times out of 100.
Anti-aliasing filters are common and further gimp sharpness.
Images are almost never displayed back 1:1 from sensor to display (nobody has 6000x4000 monitors for viewing their DSLR shots).
Copy pasting perfectly aligned pixels into a pixel grid then rotating them is taking perfection and degrading it. This is insanely visible however when you're rotating literal imperfection to begin with the actual difference isn't anywhere near as visible.
