3 ms·

It's somewhat odd to suddenly see lots of excitement in the 'continuous-tone still images' space, and I can't help but think that on some level, JPEG the group

It's somewhat odd to suddenly see lots of excitement in the 'continuous-tone still images' space, and I can't help but think that on some level, JPEG the group to wants to retain relevance at a time when the world is finally looking poised to move on from using its oldest work. A still image file containing one frame of a video codec has been long theorized, pushed very aggressively since 2010 by Google as WebP (i.e. VP8-in-RIFF), and as a fully-functional thought experiment by ffmpeg's Fabrice Bellard since 2014 in BPG (HEVC-with-custom-header), the latter of which jumped the gun on HEIF before it was done. Now that HEIF is here, we have a better container encapsulating an advanced codec, and suddenly we have hardware support with premium phones generating files in this format. The barrier is HEVC's hefty licensing pool, but a parallel effort exists in AVIF [1], which takes a frame of the royalty-free, HEVC-competitor AV1 video codec, and stuffs it into a HEIF container. It seems a bit superfluous for JPEG to barge into this space right now; I'm confident they'll do technically solid, and even innovative work, but proliferating an incompatible implementation right when it looks like there's a realistic chance of a newer still-image codec replacing one that dates back to 1992 is a bit counterproductive. But perhaps I should be more optimistic. Their calls-for-proposal document [2] lays out recent competitors, admits their technical merits, but concludes that they're not in widespread use. Then, it lays out that their vision to produce a state-of-the-art format that will achieve widespread adoption, helped by a "royalty-free baseline". [1] https://news.ycombinator.com/item?id=16803363 https://news.ycombinator.com/item?id=16803363 [2] https://jpeg.org/downloads/jpegxl/jpegxl-draft_cfp.pdf https://jpeg.org/downloads/jpegxl/jpegxl-draft_cfp.pdf

vanderZwan 8y ago
> lays out recent competitors, admits their technical merits, but concludes that they're not in widespread use. So their argument is "this new technology that has just entered the market is not widely adopted, therefore our technology which does not even exist yet let alone is ready for usage is better"?
m0dest 8y ago
>A still image file containing one frame of a video codec has been long theorized The I-frame encoding in MPEG-1 was basically identical to baseline JPEG. The compressed frames can be losslessly converted to JPEG and vice-versa. So, in a sense, we were already there. Reusing the I-frame compression from a video codec is the only strategy that makes sense for a lossy still image codec. Video codecs are where the major innovation is happening, with wide participation from every player in the industry, and a large part of the innovation has been around improving I-frame efficiency. They're being hardware optimized and baked into software implementations. And even when patents are involved, still image codecs are usually able to piggyback on the video codec licenses that are already on the device.
- ksec 8y ago
  It is interesting that H.262 ( MPEG 2 ) > H.263 ( Divx ) > H.264 ( AVC ) > H.265 (HEVC ) That is 4 generation a video codec improvement, and yet as an image format, MPEG 1 based Jpeg is still doing very well compared to latest HEIF. I mean four generation of improvement and we only got 50% reduction at similar quality. Why is it still image don't scale down as well as video?
  - niftich 8y ago
    Because many of the gains of successive video codecs comes from finding more efficient ways of encoding motion, which is something you clearly can't exploit with stills.
  - derf_ 8y ago
    Because different things sitting next to each other in space are not as inherently predictable from each other as the same thing moving through time.

vanderZwan 8y ago

> lays out recent competitors, admits their technical merits, but concludes that they're not in widespread use. So their argument is "this new technology that has just entered the market is not widely adopted, therefore our technology which does not even exist yet let alone is ready for usage is better"?

m0dest 8y ago

>A still image file containing one frame of a video codec has been long theorized The I-frame encoding in MPEG-1 was basically identical to baseline JPEG. The compressed frames can be losslessly converted to JPEG and vice-versa. So, in a sense, we were already there. Reusing the I-frame compression from a video codec is the only strategy that makes sense for a lossy still image codec. Video codecs are where the major innovation is happening, with wide participation from every player in the industry, and a large part of the innovation has been around improving I-frame efficiency. They're being hardware optimized and baked into software implementations. And even when patents are involved, still image codecs are usually able to piggyback on the video codec licenses that are already on the device.

ksec 8y ago
It is interesting that H.262 ( MPEG 2 ) > H.263 ( Divx ) > H.264 ( AVC ) > H.265 (HEVC ) That is 4 generation a video codec improvement, and yet as an image format, MPEG 1 based Jpeg is still doing very well compared to latest HEIF. I mean four generation of improvement and we only got 50% reduction at similar quality. Why is it still image don't scale down as well as video?
- niftich 8y ago
  Because many of the gains of successive video codecs comes from finding more efficient ways of encoding motion, which is something you clearly can't exploit with stills.
- derf_ 8y ago
  Because different things sitting next to each other in space are not as inherently predictable from each other as the same thing moving through time.