Announcement

Collapse
No announcement yet.

Google Continues Pushing Its WebP Image Format

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Originally posted by zanny View Post
    Opus and Dalaa seem be proving great competency on Xiphs part to just toss out 30 years of cruft and figure out modern ways to get great balances of compression, latency, and decode time. I'd love to see what they could do with static images.
    They've been applying their compression to static images as a way of testing it, but I think I might ask them about it as an image format.

    Comment


    • #17
      Originally posted by bondTongue View Post
      Is the BBC not able to work pngout, or zopfli into their workflow? I wonder, because using pngout on your image saves 3%, and zopfli saves 6% both losslessly. They are missing out on a 50KB savings , simply because they can't run a command line program? Here is a losslessly compressed version of your banner http://i.imgur.com/NUhAsrA.png I made it completely automatically.
      I just saw this message, but no, it just doesn't make sense to install additional unofficial software onto multiple servers so that we can reduce the image sizes slightly, especially since we already have access to WebP as a part of PHP. The problem we want to solve isn't our bandwidth costs, but the cost in website responsiveness from having 5MB of banner images and lots of people having poor internet connections. 50KB isn't going to cut it.

      I should also point out that I no longer work for R&B Creative who are responsible for the BBCWW websites, so I won't actually be doing anything

      Comment


      • #18
        Originally posted by NomadDemon View Post
        give me jpeg with transparency, and i will use it again
        It's called JNG. At its peak it had more support than webp has now, though now it's mostly been removed since adoption didn't take up.

        Comment


        • #19
          Originally posted by tomato View Post
          why it will never support them? (honest question)
          OK, I was wrong about arithmetic which is supported by libjpeg-turbo but rarely anybody enables it (there were also patent problems with arithmetic but I think all patents have now expired).. Even so, when you use jpeg with arithmetic you never know if the user will be able to see it or not - which makes this really useless.

          As for other features see http://www.libjpeg-turbo.org/About/Jpeg-9. In short.. they are non-standard extensions and rejected by ITU-T.

          Comment


          • #20
            Originally posted by quikee View Post
            OK, I was wrong about arithmetic which is supported by libjpeg-turbo but rarely anybody enables it (there were also patent problems with arithmetic but I think all patents have now expired)..
            It's likely disabled in many implementations for historic patents reasons, unfortunately. As you noticed, the patent expired (many years ago). Arithmetic coding is part of the JPEG standard. Toggling the switch to enable it ought to be much simpler than introducing a completely new format!

            Even so, when you use jpeg with arithmetic you never know if the user will be able to see it or not - which makes this really useless.
            WebP already has the same issue, it's not fully backwards compatible and some features are optional.

            As for other features see http://www.libjpeg-turbo.org/About/Jpeg-9. In short.. they are non-standard extensions and rejected by ITU-T.
            The older JPEG-LS format extension is standard (ITU) and popular in some domains. The SmartScale stuff is rather strange, yeah, but I wasn't talking about that.

            Comment


            • #21
              Originally posted by psychoticmeow View Post
              It even works for normal screens, because instead of displaying the image at 100% of its width you display it at 50% and let the browser scale it. The end result is pretty much the same, instead of serving a 100KB JPEG you're now serving a 60KB WebP.
              I see that Google logo works that way, it's a 538x190 png that is displayed at 50% (269x95) on their front page.

              So I tried to experiment a little with it.

              14022 bytes: original logo11w.png
              12644 bytes: lossless webp (cwebp -lossless -m 6)
              12558 bytes: zopfli optimized png (100 iterations)

              It looks like 10% improvement either way.

              Comment


              • #22
                So what's wrong with SVGs that people don't use those for dynamically sized icons?

                Is there another vector image format that isn't a textual, xml based bloat magnet?

                Comment


                • #23
                  Originally posted by zanny View Post
                  So what's wrong with SVGs that people don't use those for dynamically sized icons?
                  You can't SVG-ise a photo without ending up with a raster. So for many things SVG is a good idea, but not for things that are very complex.
                  There are some pretty amazing complex SVGs, but at that point it gets cheaper to just put a raster, for instance (size parity only at 2700x1800): https://commons.wikimedia.org/wiki/Fileojikko2.3.svg
                  Last edited by GreatEmerald; 03-24-2014, 10:48 AM. Reason: Turn off smilies

                  Comment


                  • #24
                    Originally posted by psychoticmeow View Post
                    They've been applying their compression to static images as a way of testing it, but I think I might ask them about it as an image format.
                    I, too, I think that wrapping their prediction algorithms into an actual image format could provide some definitely interesting results.
                    (Probably even better then WebP, and in the range of what top-of-the-line wavelet compressors produce).

                    Comment

                    Working...
                    X