Slashdot Mirror
Bellard Creates New Image Format To Replace JPEG
An anonymous reader writes Fabrice Bellard (creator of FFMPEG, QEMU, JSLinux...) proposes a new image format that could replace JPEG : BPG. For the same quality, files are about half the size of their JPEG equivalents. He released libbpg (with source) as well as a JS decompressor, and set up a demo including the famous Lena image.
Because the new half-the-size JPEG files wouldn't work with old JPEG editors/viewers.
While I'm generally also skeptical, when JPEG2000 was released, decoding images in JavaScript wasn't an option. As such, there's not really any barrier for individual websites to switch, if they're heavily image-driven.
A bigger roadblock might be that these days, bandwidth (and storage) is cheap, and so savings in image size are less relevant than they used to be.
Same-origin policy is a nightmare for use with CDNs. I really wanted to use WebP for image handling for the application I'm working on, but Firefox adamantly refuses to merge a patch adding WebP support, and the JavaScript shim can only access the images if they're pulled of the same host. Images loaded from a CDN aren't accessible to the JS decoder.
1/2 the size of jpg for equivalent quality. I'm sold.
As soon as Photoshop and Firefox/Chrome start supporting it I can see widespread adoption.
The below site offers a better comparison interface than the Lena image link from the post. Drag your mouse across the image to see the effect:
http://xooyoozoo.github.io/yol...
It's worth noting the demo page is using JavaScript decoder to display the images; so it seems more than feasible to transition to the new format by first just having JavaScript decoder do the displaying on image-intensive sites. Still, I have to agree that especially with todays website-bloat and bandwidth "Another new format to pack your images even smaller!" isn't likely to fly. If the headline was much better quality, maybe, but it's not immediately clear to me that this is in any way better than just using higher quality/size JPEG. (Although as hinted, image-intensive sites who pay for their own bandwidth surely disagree!)
From Wikipedia
[24] http://www.mpegla.com/main/pro... (PDF)
Except that this test image has just a face and part of a shoulder, without any naughty bits. Not even erotic at all.
It's a good test image because it catches both distortions of detail and color damage to areas with a gentle gradient.
The creatures outside looked from Alt-Right to Antifa; but already it was impossible to say which was which.
Bandwidth still matters for mobile, so smaller images of the same quality are quite welcome on mobile sites and apps.
Given that the developing world is likely to get online via wireless solutions, bandwidth is going to matter for a lot of people for a long time to come.
Yesterday it worked; today it is not working; Windows is like that...
Because JPEG 2000, JPEG XR, WebP and others don't work in browsers without specifically added support, BGP does (via javascript).
The creatures outside looked from Alt-Right to Antifa; but already it was impossible to say which was which.
If you have the ability to change the server configuration, wouldn't CORS work? "Access-Control-Allow-Origin: *"
From the web site "BPG natively supports 8 to 14 bits per channel," which is a huge advantage. 8 bits is more of a straight-jacket than people realise and this offers a more portable way for people to pass around high bit-depth issues than camera raw files (proprietary things inside) or TIFF (a complex container format prone to cross-platform issues and poor compression).
JPEGs do support alpha channels, browsers just might not render them properly, but same goes with JPEGs with CMYK colorspace.
- Raynet --> .
Note that, according to the BPG web site, "An alpha channel is supported" so BPG has transparency.
How are we going to pronounce this thing? "Bee-Peg" I suppose since "Bee-Pee-Gee" doesn't roll off the tongue.
Looks good.
Because it has never been done before? http://bstring.sourceforge.net...
It really is political correctness gone off the deep end when anyone even *cares* that the standard Lena image is cropped from Playboy or that it depicts a female. Who cares! It's an image. The standard argument is that it's "sexualized" and thus offends the sensitivities of some feminists. I call bullshit.
First of all, nobody uses this because it's sexualized; much better free porn is widely available to anyone on the Internet anyways. People use it because it was once used historically for image comparison, and everyone since has used the same image so that their improvements to image compression algorithms can be directly compared to results published in older papers.
To even call it sexualized is borderline lunacy unless you're from an isolated religious cult or are living 200 years in the past. There's like, a whole bare shoulder in that photo. If that's going to offend you, you should really try to avoid ever going to a beach or walking down a public sidewalk, lest you run into so many bare ankles and shoulders that you might succumb to a panic attack.
Even if we were to accept that there's some sexualization indirectly through the context that it was cropped from Playboy (which none of the image comparison papers ever make a big deal of to begin with), why does a sexualized image offend, and why does it only offend when it's female? I'm a standard geeky male with a bit of a pot belly. Am I supposed to be enraged about body-image issues or offended at the world every time Calvin Klein interrupts my TV viewing by showing an ad with a hunky male model in boxers? Grow the fuck up, feminists.
Barney Stinson settled the pronunciation many years ago on How I Met Your Mother: it's "bee-peg". Although this format probably won't be used exclusively for pictures of boobs.
The main issue that is going to hold back adoption is the use of HEVC/H.265 as compression codec. While dedicated hardware is not needed to decompress the images in a timely manner, it also means that no licensing fees have been paid to the MPEG LA. Since the format is patent encumbered, I can't see this taking off unless the patent pool decides to give out a royalty free license for still images. Bellard himself assumes that most hardware will come with said codec licensed and built in but that does not include old hardware or even current hardware that is not being shipped with it. Barely anything ships with H.265 support other than the iPhone 6 and a couple of Mediatek SoCs.
So what constitutes a "product" when it comes to software? Say Mozilla implement this new format for Firefox, does that mean Mozilla have to pay $0.20 to MPEG LA every time someone downloads a copy of Firefox?
The demo website linked in the story sent a BPG decoder implemented Javascript to our browsers. So does that mean Bellard owe MPEG LA a metric shit-ton of money now?
But JPEG2000 was absolutely crawling with patents like maggots and worms writhing through the very core of its being. If that didn't put everyone off then I don't know what would? Certainly ruined my lunch.
DJVU was another contender but it just happened to be tagged on to a PDF-like docuemnt format and not widely known as just an image format.
Finally, anything that was not (properly) supported by Internet Explorer ten years ago was a dead duck. And Microsoft and Apple actively snub any open format if they can get away with (like Vorbis, WebM etc).
I know it's a good image for this purpose, it's just the origin that is really sexist. I don't mind, but I can understand how some of our female colleagues would find it inappropriate - ranging from immature to degrading, depending on where in their cycle they are right now...
I don't think we should compare BPG with JPEG, since it is very outdated. I wonder how it stacks against WebP - does it also support animation? Better compression? Licenses? Faster encoding/decoding? Browser manufacturer support? I'm all for making web more optimal, because you can never have "fast-enough" bandwidth, especially on a mobile device in bad connection area, but lets compare similar things.
I realize that this is Slashdot and we have a great tradition of not RTFA, but given that this is about an image format you could at least go LATFP (Look At The Fucking Pictures). It's also an impressive display of how well image deciding using JavaScript works (but then, this is the guy who wrote an entire x86 emulator capable of running Linux using JS, and even made it work on IE; I have no doubt as to the man's skill in that realm).
Link for image format and quality comparisons: http://xooyoozoo.github.io/yol...
Link for info about the image format and links to more comparisons: http://bellard.org/bpg/
There's no place I could be, since I've found Serenity...
Boosting the signal, for those who don't read ACs:
CORS (Cross-Origin Resource Sharing) is explicitly intended to support things like CDNs. It lets you make cross-domain XHRs (and access the responses), so the JavaScript-based decoder will work perfectly. It adds minimal additional bandwidth requirement over a standard cross-domain GET (one short extra header on request, a couple on response), is supported on all mainstream browsers, and is much more secure that stupid hacks like JSON-P (though that would work here too, if for some reason you wanted to live in last decade's terrible work-arounds for same-origin policy).
http://en.wikipedia.org/wiki/C...
There's no place I could be, since I've found Serenity...
If they have to be told what the image is from to get upset, to paraphrase Steve Jobs, they're getting offended wrong. The image as it is now does not objectify women any more than images that run in modern newspapers. Once it's been cropped to this level, it's literally lost all value that made it sexist to start with. If you can tell them instead it's from some old ladies fashion magazine and they're suddenly okay with it, I'd have to say you proved my point.
Yes and yes, respectively. Though for the latter they probably won't bother.
This is just a terrible idea.
Not so much a "limit" as much as a complete showstopper.
While my inclination is towards BPG, the argument could be made that it would be superior to implement a javascript decoder for those other file formats, if they provided better quality at lower file sizes...
"Go to CNN [for a] spell-checked, fact-checked summary" -- CmdrTaco
I was pleased to see that Matplotlib switch to Grace Hopper as their test image. http://matplotlib.org/examples...
I'm not sure how you can argue that after looking at the pictures in the link. It's clearly superior to JPG, because *everyone* can see the JPG artifacts. You only tend to notice the artifacts with BPG if you're comparing to a high quality picture or the original, or else looking really hard. It seems similar in principle to good audio compression that saves space by removing details the human ear is unlikely to miss.
It's too bad, because we really could have used this years ago while we were still on dialup - it would have saved us from seeing many beautiful images compressed all to hell. Yes, bandwidth matters to some degree nowadays, but not nearly as much as it used to. This format will, unfortunately, probably get little traction for one reason. JPG is here and it's "good enough". Audiophiles chafe at MP3 as well. Technically speaking, Ogg Vorbis was a superior format in nearly every way, but it's widely ignored in favor of MP3, which is "good enough". There's a small movement with FLAC and hi resolution sound, but most people can't hear or don't care about the difference. It will probably be the same for this.
Who knows... maybe I'll be proven wrong. It would help if the browser makers actually got behind it early and supported it fully - PNG suffered poor adoption because IE lagged so far behind with support for many years. Adobe, Corel, and other makers of image software will also need to offer native support as well. A format is worthless unless people are actually using it.
Irony: Agile development has too much intertia to be abandoned now.
I was hoping for the complete Lena. When the image popped up, in the 1970s, sure that the larger parts of the image were cut off for indecency.
But in 2014, I think this is no topic any longer.
A new coding algorithm could as well have come with a new perspective on morals.
And given us something NSFW, to look at in the workplace!
Note that, according to the BPG web site, "An alpha channel is supported" so BPG has transparency.
Now, that should have been the headline instead of saving a few poxy KB.
Wanna buy a shirt?
https://www.redbubble.com/people/stealthfinger/shop?asc=u
JPEG2000 in Javascript
WebP in Javascript
Another WebP in Javascript
"Go to CNN [for a] spell-checked, fact-checked summary" -- CmdrTaco
The problem here is that H.265 and by extension BPG are heavily patent-encumbered. These are not just latent patents but patents that the H.265 contributors are using for a revenue stream.
Bellard suggests "just use the licensed hardware decoder you probably already have" but a) that doesn't make technical sense in lots of cases and b) most people don't, in fact, have such a thing currently and c) the encoding situation is even worse.
There's more than one way to compress images though, some vastly different to others. BPM is working on the original image and compressing better than JPEG. As for whether that loses more data, that's not a given - it is possible for a different algorithm to compress to less data than JPEG but retain more information.
JPEG is (barring the possibility of some lossless mode that looks very little like JPEG except for a few metadata fields; but is technically part of the spec, not sure if we have one of those) indeed compressed; but it's lossy compression and lossy compression is an area where there is actually a reasonable amount of ongoing development.
This isn't to say that lossless compression is a trivial problem, or that there have been absolutely no improvements; but the 'by definition, it isn't lossless unless applying the decompression operation to the result of the compression operation produces something identical to the input' criterion makes it much easier to let the mathematicians and computer scientists work out the limits of the possible.
With lossy compression, there aren't any formal limits, which leaves the field much more open to solutions that rely on following the strong and weak points of human perception(visual in this case, auditory in other cases, visual/motion related in others), which leads to much greater complexity and diversity.
A bigger roadblock might be that these days, bandwidth (and storage) is cheap, and so savings in image size are less relevant than they used to be.
Bandwidth isn't cheap on mobile data networks. On the other hand, requiring phones to execute battery sapping image decompression in Javascript is hardly a great idea either.
With 'mobile' there are really two considerations: One is the fact that 'mobile' (even if the fault is, in fact, with shitty backhaul) is going to be fairly slow in emerging markets. Two, relevant even in wealthy developed nations without asshole oligopoly telcos, is the fact that mobile devices are brutally power constrained, and RF chatter isn't exactly cheap in energy terms. The faster you get the data you need and shut down, or move to a slower, lower power mode, the radios, the happier your battery will be.
With mains power it matters less (electricity isn't free; but a few extra dollars a month is far less annoying than having your battery keel over dead at a bad time); but barring exciting breakthroughs in battery chemistry or design, basically all the savings are going to have to come from the device side.
Because it has a catchier name.
I mean, "JPEG 2000"? Seriously? That is -so- 20th century.
You and your friends who can get decent bandwidth, can afford decent smartphones and who can afford to just throw down an additional 2 euro a month for said bandwidth are, you may be surprised to hear, not representative of everyone. For example, the average internet connection speed in Algeria is about 0.94Mb/s. I'm pretty sure most people there are also not wandering around with the latest LTE enabled phone either.
That, and potentially somebody popping out of the woodwork with a patent claim, however dodgy.
For a free format without any sponsors(either altruistic or, as with WebM, probably doing it as a tactical move to improve their negotiating position if the MPEG LA ever decided to shake down youtube for more than token fees for H264), it doesn't take a terribly plausible patent claim to be plausible enough that it'll be hard to find somebody to go slug it out in court.
As much as certain formats are...getting rather elderly...it certainly is handy to be assured through sheer age that any patents that might apply to a given format have expired.
PNG is used extensively in Apple products. It's the standard format for non-compressed images in iOS apps, along with jpg for compressed images. Apple recommends using png for user interface elements, and jpg for pictures. Which makes sense, since jpg can compress a picture to 20% of the size with very few artifacts. Size does matter for mobile apps. And I wish people would realize that it matters for servers, too. Yes, available bandwidth is enormous these days. But if your server is serving pages that are twice the size and your audience is large, you're going to need a bigger server room and use considerably more energy. Servers are on their way to becoming the biggest power hogs on the planet. Smaller images mean less hard disks and less data pipes.
If you're running a server for a big company (say, Google or FaceBook) and every image is only half as big, that means a huge reduction in the number of servers you need, power consumption, etc. Less congestion on the internet, more responsive servers, less wasted energy, etc...
I imagine you also have a car that guzzles up twice as much gas as other cars, but who cares since you can afford it?
I'm a videogame programmer, and Ogg Vorbis is actually a very popular format for game audio. It's not only license free, but it supports multichannel audio and seamless sample-accurate looping, which standard MP3 can't do. It was great for videogame companies, but did little to really promote the file format itself. So, sure, the fact that we have usable reference libraries means anyone can add support to their products, but I don't think that makes much of a difference, unfortunately.
Don't get me wrong - I think it's a great format (obviously technically superior) and would love to see it succeed. You say that if the format "becomes a standard implemented by browsers and major graphics tools, it will get adopted". Well, sure, but that's sort of the hard part, right?
Irony: Agile development has too much intertia to be abandoned now.
In this industry, there's no such thing as a "replacement," it's "just another competing format." None of the old formats ever dies, all we ever get is more new formats, all of which need to be supported, ultimately making everything more complicated. I'm not saying we shouldn't advance... but the belief that some new format you create will replace something instead of muddying the existing pool of formats is laughable. related xdcd. (yes, I know it's "standards" and not "formats," but the result is the same)
Stupid sexy Flanders.
You don't have to wait for someone to pop out of the woodworks. BPG is nothing but a still frame of HEVC video which is patented up the ass. Bellard and other open source video authors are accustomed to ignoring the patent situation because they don't really have a choice if you want to be interoperable, but that isn't an excuse for creating patent problems in a field where there are already widespread royalty free standards (JPEG, PNG).
The first question to come to my mind is who has the patents on this animal, and how long will it be before the lawsuits begin? They'll probably wait until the new format is firmly established on the Internet before springing the "gotcha" on folks.
"My country, right or wrong; if right, to be kept right; and if wrong, to be set right." --Senator Carl Schurz (1872)
How are we going to pronounce this thing? "Bee-Peg" I suppose since "Bee-Pee-Gee" doesn't roll off the tongue.
;)
Oh come on, you literalist! You should clearly pronounce it with a soft "B".
Double the image quality for the same bandwidth. I want this format supported in all browsers yesterday already.
If you don't risk failure you don't risk success.
It's a soft G: "Bee-Pej".
-- Don't Tase me, bro!
HEVC, ouch, that'll sting pretty brutally.
The one interesting quirk about being directly based on a video codec, though, is that a variety of devices (eg. SoCs with fixed function decode blocks, computers with OS vendors that pay the appropriate tithes, etc.) might be 'automatically' licensed for the special-case use of treating a single HEVC frame as a still image by virtue of being suitably licensed for HEVC playback. It'd be amusing to see whether that argument is accepted, or whether 'single frame playback' will mysteriously require a different set of licenses.
Not a patent minefield I want to wade into, though, though I do very much appreciate the work of those who ensure that we have access to solid implementations of these standards, even if it is far beyond their power to untangle the patent nightmare.
Servers always work to reduce bandwidth usage. Bandwidth is expensive when you're talking thousands of users.
Smaller images means faster transfer and faster load times, especially for mobile.
Just look at all the efforts put into bundling/compression/etc. Some companies go as far as reducing all their CSS class names to 3 or less characters. These have different purposes though not always directly related to bandwidth reduction. Bundling is more about reducing the number of HTTP requests than reducing bandwidth though, since it bundles multiple requests for CSS/JS into single request for each, because each HTTP request consumes server resources.
Usage is larger than it used to be as well. Now vast majority of people have a computer in their pocket at all times and access internet much more frequently than the age of desktops, when there was one computer per family accessed intermittently.
Many mobile data connections have lower bandwidth than traditional ground connections, although a few are faster.
As for as harddrives, the pervasiveness of digital cameras being on every phone in many pockets, means a tremendous increase in # pictures being taken. Storage on phones is higher $/gb than hard drives. Usually these make their way onto a server such as instagram or facebook, who each would be interested in reducing storage size, as the $/gb is high when you consider that data likely has at least two forms of redundancy.
...and the downscaled, recompressed version will probably be the only version kept in the archive.
Got them moderator blues I blieve I walk out the do', With these mod-points I been gettin', I 'most never post no mo'
There is a reason why JPEG is blocky. The blocky nature of the encoding preserves details better.
BPG blurs everything heavily. Small details and fine textures literally disappear.(*)
JPEG is definitely outdated and web could gain from a worthy replacement. But BPG IMO doesn't appear to be "it".
(*) I wonder how JPEG would fare on the images, decoded from BPG. Since fine details are removed by BPG, the JPEG would be smaller too.
All hope abandon ye who enter here.
Remember that in many emerging markets, Blackberry is the way most of the population access the web.
Cool. A time traveler. What year are you from?
Faster! Faster! Faster would be better!
Is BPG using some form of VQ (vector quantization)? What are the key tech improvements over JPEG?
Um... would you like a beer, perchance?
When someone says, "Any fool can see
Is it any different/better than already existing PGF file format?
https://en.wikipedia.org/wiki/...