File talk:Michelangelo's David - Floyd-Steinberg.png

From Wikimedia Commons, the free media repository
Jump to navigation Jump to search

This image is not dithering the source image properly

[edit]

@AtomCrusher I would like to change this file since I recently found that it might not be a proper Floyd-Steinberg dithering of File:Michelangelo's David - 63 grijswaarden.png. I was writing C code to dither a ppm image and after comparing the output with File:Michelangelo's David - 63 grijswaarden.png, I found it had some differences. I first assumed something was wrong with my code, but just to make sure, I went to https://ditherit.com/ to compare outputs. To my surprise, my code's output and ditherit.com's output were identical. I also compared with the output of https://www.luxa.org/image/dither, https://29a.ch/ditherlicious/, https://shihn.ca/posts/2020/dithering/, and https://doodad.dev/dither-me-this/. luxa.org produced the same output as my code, however all others seemed to produced unique outputs. I suspect this is due to filters for ditherlicious and doodad.dev, but I cannot explain it for shihn.ca.

I wanted to ask permission first since I am relatively new to this type of editing and wasn't sure if this would violate the rules on Commons:Overwriting existing files. @Gerbrant has not logged on since 2008, so I thought I should mention @AtomCrusher instead. BlastKast (talk) 04:04, 2 December 2022 (UTC)[reply]

The threshold for the dither could be different, or some intermediate value rounded, hence the differences. If you do want to change it then go ahead - all I did was recompress the image - and you can always add another file in Commons with a different filename if you're worried. Though I'd suggest an overwrite is appropriate as a "correction" if you feel the need. -- AtomCrusher (talk) 01:15, 6 December 2022 (UTC)[reply]
Thanks for the response! My concern was that if somebody applied the algorithm to the source image according to what is defined on the Floyd–Steinberg dithering article, it would not create the same output. I am aware there can be slight differences in the algorithm, but I felt that the image should be an accurate example to what is listed. At the moment I am somewhat busy, so I'll look over it again and get to updating it in a couple days, if that is okay. BlastKast (talk) 05:00, 6 December 2022 (UTC)[reply]