Like divix format in the video world, now what we need is an ultra-high-compression & quality image file format

[compuser]

Does such a file format exist to date that can rival the compresson as well as the quality of the equivalent new formats in the world o' video and audio (ie. divix, mp3, etc.)?

Sure, JP(E)G is a great format especially at 0% compression setting, but it goes hand in hand with capabilities of mpeg. We all thought mpeg was impressive enough and quality enough... till the new kids on the block came about. I can't wait till such a format for image compression and quality retention is created! Just imagine 4-6MP crisp clear digicam/1200dpi scanner images being stored on a floppy disk and looking like a TIFF!! It's just a matter of time....

 

[Sahakiel]

Awhile ago, I heard of a format which reduced image sizes based on geometric shapes. Supposedly highly compressed, but highly computationally intensive. So much so it was impractical at the time.

 

[sxr7171]

I've never seen a high-bitrate MPEG-4 video, is it that much better than MPEG-2?

 

[amok]

It's slightly lower quality. The great thing about divx is that it is near dvd quality at 1/10 the file size.

 

[glugglug]

RAR multimedia compression on an uncompressed bitmap gets about 30 times better ratio than JPEG at a 90% quality setting.

And the RAR compression is non-lossy.

 

[CTho9305]

Originally posted by: glugglug
RAR multimedia compression on an uncompressed bitmap gets about 30 times better ratio than JPEG at a 90% quality setting.

And the RAR compression is non-lossy.

Do you have an example of this you could post online?

 

[DRGrim]

Originally posted by: compuser
Does such a file format exist to date that can rival the compresson as well as the quality of the equivalent new formats in the world o' video and audio (ie. divix, mp3, etc.)?

Actually, i believe that mp3 is fairly old. There are newer and better codecs available, such as Ogg Vorbis and Monkey's Audio.

As for your question, JPEG2000 looks promising. It can do high quality-low filesize compression, lossless compression, and other neat things, such as the ability for the server to resize the image before it is sent out (which lets you get rid of multiple files for thumbnails, et cetera).

EDIT: Here is an example of the difference between JPEG and JPEG2000

 

[Shalmanese]

Sure, these guys in the Netherlands made this super-ultra-woser incredible compression format that could fit an hour of video onto a floppy, apparently, its meant to be out some time 6 months ago /Sarcasm.

One of the things that allowed divX to have such a great compression ratio was that it based its assumption on thr fact that this frame will be a lot like the last frame. Images dont have that luxury.

 

[Chu]

>> Like divix format in the video world, now what we need is an ultra-high-compression & quality image file format

Divx (and other mpeg4 encoders) attain their ultra-high compression via processes that do not work for still images, such as motion prediction and B frames. For keyframes (i.e. the frames in your video that are essentially still pictures), the compression isn't much better then jpeg or png.

-Chu

 

[Nefrodite]

Originally posted by: sxr7171
I've never seen a high-bitrate MPEG-4 video, is it that much better than MPEG-2?

nope, divx pulls off decent images at low bitrates, but once you reach dvd level bitrate(which mpeg4 doesn't seem to be optimized for,it simply won't encode that high) mpeg2 shines. mpeg4 seems to have a 256color effect sorta, high compression jpg artifacts is best i can describe it.

 

[Chu]

>>

nope, divx pulls off decent images at low bitrates, but once you reach dvd level bitrate(which mpeg4 doesn't seem to be optimized for,it simply won't encode that high) mpeg2 shines. mpeg4 seems to have a 256color effect sorta, high compression jpg artifacts is best i can describe it.

<<

MPEG4 is superior in every way to MPEG2 except resources needed to decode it. This is the fault of the encoders, not the format.

-Chu

 

[compuser]

Interesting! I'd never heard of JPEG2000. However, I wish they would have done a better job of marketing the file format by putting up images of the SAME quality as the source image (which should be a TIFF or a 0% loss JPEG or something similar) and then show the highest compression possible while not revealing any visible flaws in the new JPEG2000 image.

Criticism aside, it does look promising. I whish it were free (at least the site doesn't hint that it is). Anyway, as with any technology, it comes with a premium.

As for the comment about it being difficult-to-impossible to use MPEG algorithm (prediction based) to static, single-frame image; I agree. However, I am sure the powers that be, out there are therefore circumventing this issue in their algorithms.

I've always wondered if this idea is already being used: an algorithm that will store the information of say 568 adjacent pixels of same color in some compressed form by saying defining that a polygon with these n corners encompasses that color. This poly might not cover the entire equi-color area but can compress the data quite a bit by finding the largest poly that can fit in this (odd) shape of 568 pixelsa polygonal area... something to that effect.

anyway, just my 2c worth. Again, "It's just a matter of time..." I believe, before such a open source format becomes available.

EDIT:

Doh! Didn't read the 2nd post by Sahakiel. I guess, I'm just reinventing the wheel here then. There are just too many Einsteins out there, dime-a-dozen.

 

[DRGrim]

Here is another interesting technology, although far from mature. Fractal Image Compression. There are better sites, I just don't have the time to look for them right now.

 

[CQuinn]

MPEG4 is superior in every way to MPEG2 except resources needed to decode it. This is the fault of the encoders, not the format.

MPEG-4 is a streaming media extension of MPEG-2, while removing many of the features of MPEG-2 that were
not considered viable for a streaming dataset.

It also does reduce the color palette used for encoding from the MPEG-2 standard, so no encoder -
however much they are improving, would be able to match the exact quality of a DVD source after converting to
DivX, Xvid, WMA or any other MPEG-4 based codecs. MPEG-4 encoding does take more resources to
do similar quality to an MPEG2 encoding. And it will scale itself to the system it is being played back on, so
even in decoding the quality can be less than what you would get with a similar DVD stream.

MPEG4 is a different way of encoding video for different uses than MPEG-2, that alone does not make it superior.
It (just like MP3) is "good enough" for most people to not care about the difference most of the time.

 

[crypticlogin]

However, I wish they would have done a better job of marketing the file format by putting up images of the SAME quality as the source image (which should be a TIFF or a 0% loss JPEG or something similar) and then show the highest compression possible while not revealing any visible flaws in the new JPEG2000 image.

Flaws and quality grading are problematic measures since quality will always be subjective and dependent on the image content. e.g. In some formats, the borders between two distinct colors are pixellated while in others, smoother. Which artifact do you want less of, or more appropriately phrased, would mind the least? Then there's color gradients and how smoothness (large blocks of a single color) or pixellation (more color range but speckled) could ruin the image. You've probably seen this in pictures of a crystal clear sky.

Personally, I'd like to see a standard, uncompressed test image that everyone uses to compare schemes. That way, there's no fudging the "our scheme is the best -- just look at this proprietary image we choose after days of looking for one that works well for us!" process. In the image coompression sense, it'd gauge compression ratio, artifacts, and overall image quality. Isn't there some test pattern that photographers use to gauge lighting, focus, clarity, etc? Or is that proprietary as well?