FLIF-hub / FLIF

Free Lossless Image Format
Other
3.72k stars 229 forks source link

Visually, how does MANIAC understand images? #525

Open alan2here opened 5 years ago

alan2here commented 5 years ago

Bitmaps store pixels in a very simple way, it wouldn't be surprising if a change to a part of a bitmap file created from a photo was almost imperceptible where just one pixel changed colour.

For highly compressed lossy JPEGs a tree/fractal of squares is clearly visible, with messy seems between regions. Probably less apparent is that JPEG is apparently also makes heavy use of the frequency domain.

With MPEG video removing certain frames turns videos into striking visualisations of how keen the format is to understand inter-frame movement as spacially displacing pixels/regions from the previous frame.

Projecting into the future I could imagine a video format, perhaps based on neural networks where a minor bug in a file could have the TV detective believably accusing the wrong suspect of the murder at the end of the show, having assembled the clues differently to reach a different conclusion.


Are there any such straightforward visual indications of what properties FLIF exploits to convert the 0s and 1s into photos, drawings etc…?