honestly i’m really skeptical of all these video compression algorithms trying to be image compression algorithms. maybe i shouldn’t be talking because mao quote but the image algorithm designers never had the advantage of deltas between frames. i assume there’s some patent thing stopping apple from using jpeg-xl instead of heif (which is avif but for hevc instead of av1) but since i can’t find any benchmarks i’m just speculating there. but it’s all speculation since i don’t really understand compression all that well.
EDIT: found a paper by google suggesting heif and jpeg-xl have similar perception scores with jpeg-xl giving a slightly lower size. however, dark spots may have slightly worse perception with jpeg-xl.
it makes sense yeah but most of the space savings is from keeping one frame and then saving the next several frames as a series of just deltas. plus there’s stuff you notice when it’s moving but not so much when it’s still. between that and the jpeg-xl benchmarks i’m comfortable speculating that there’s probably maybe something to the idea of having separate lossy image and video codecs.
you’re right but even then i kinda think there are different things to optimize for. i found a benchmark i edited into my original post but honestly who the hell knows
honestly i’m really skeptical of all these video compression algorithms trying to be image compression algorithms. maybe i shouldn’t be talking because mao quote but the image algorithm designers never had the advantage of deltas between frames. i assume there’s some patent thing stopping apple from using jpeg-xl instead of heif (which is avif but for hevc instead of av1) but since i can’t find any benchmarks i’m just speculating there. but it’s all speculation since i don’t really understand compression all that well.
EDIT: found a paper by google suggesting heif and jpeg-xl have similar perception scores with jpeg-xl giving a slightly lower size. however, dark spots may have slightly worse perception with jpeg-xl.
Kinda makes sense that storing image would be a subset of storing videos.
it makes sense yeah but most of the space savings is from keeping one frame and then saving the next several frames as a series of just deltas. plus there’s stuff you notice when it’s moving but not so much when it’s still. between that and the jpeg-xl benchmarks i’m comfortable speculating that there’s probably maybe something to the idea of having separate lossy image and video codecs.
You are still storing a key frame every couple of seconds or so
you’re right but even then i kinda think there are different things to optimize for. i found a benchmark i edited into my original post but honestly who the hell knows
To me it’s honestly a lot of
And the results in data compression aren’t different enough to use anything with less wide software support.