3 ms·

JPEG 2000: The Better Alternative to JPEG That Never Made It Big

18 points by artsandsci 10y ago

skrowl 10y ago
When can we expect the same article about WebP?
- tehlike 10y ago
  I made a good slice of mobile app ads use webp, it provides significant bw savings over jpeg, and i am not sure if it is going anywhere. Ps: i work for google.
theandrewbailey 10y ago
I think this is a bad article. I was expecting a more technical piece, but it only gives a few high level reasons why it never got used widely. And wasn't legal requirements a big reason? It mentions that backwards incompatibility was a big reason. Why would you design an obviously better media format, then ruin it by a burdensome needless requirement? Has any widely used media format been backwards compatible? (I can't think of any)
dmitrygr 10y ago
Missing main reason: licensing
bnolsen 10y ago
jpeg2k is far far more complex, hard to implement, is far more computationally complex. The quality of the results arguably aren't that much better. Revisiting the encoding of the DCT coefficients shows tremendous gains can be had there. After digging into jpeg2k I was never a big fan. IMHO for image compression there's a balance between algorithmic simplicity, computational efficiency and image quality. I'm looking for an article some months back about DCT coding sans huffman, but here's a quick page on how using arithmetic coding in the JPEG standard cuts jpeg file sizes down 5-12% http://www.rw-designer.com/entry/1311 http://www.rw-designer.com/entry/1311 Okay one DCT based recompressor is "lepton" which is 20% smaller than jpeg with the identical DCT quality: https://blogs.dropbox.com/tech/2016/07/lepton-image-compression-saving-22-losslessly-from-images-at-15mbs/ https://blogs.dropbox.com/tech/2016/07/lepton-image-compress...
niftich 10y ago
Other commenters are right, licensing was a big factor, because the 1990s were marred by a few high-profile events where patents on popular formats were suddenly enforced [1]. These events led to the PNG format being developed as an alternative to GIF, and Vorbis developed as an alternative to MP3. By the time JPEG 2000 rolled around, wariness of patents was well-established. On the technical side, JP2 produced unsightly ringing distortions much more noticeably at medium compression ratios than JPEG did while also blurring detail, while JPEG would only experience ringing around sharp transitions, and would get blocking artifacts elsewhere. For some people, JPEG's blocking would impact a sort of 'fake sharpness' in the image, while JP2 would produce a blurry smudge. Dark Shikari, longtime dev on x264, wrote about this phenomenon [3], which also adversely affected early H.264 encoders and made them look worse compared to ASP (like Xvid), despite former being an objectively better format. [1] https://en.wikipedia.org/wiki/GIF#Unisys_and_LZW_patent_enforcement https://en.wikipedia.org/wiki/GIF#Unisys_and_LZW_patent_enfo... [2] https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and_legislation https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and... [3] http://web.archive.org/web/20141124213159/http://x264dev.multimedia.cx/archives/317#more-317 http://web.archive.org/web/20141124213159/http://x264dev.mul...

18 points by artsandsci 10y ago

skrowl 10y ago

When can we expect the same article about WebP?

tehlike 10y ago
I made a good slice of mobile app ads use webp, it provides significant bw savings over jpeg, and i am not sure if it is going anywhere. Ps: i work for google.

theandrewbailey 10y ago

I think this is a bad article. I was expecting a more technical piece, but it only gives a few high level reasons why it never got used widely. And wasn't legal requirements a big reason? It mentions that backwards incompatibility was a big reason. Why would you design an obviously better media format, then ruin it by a burdensome needless requirement? Has any widely used media format been backwards compatible? (I can't think of any)

dmitrygr 10y ago

Missing main reason: licensing

bnolsen 10y ago

jpeg2k is far far more complex, hard to implement, is far more computationally complex. The quality of the results arguably aren't that much better. Revisiting the encoding of the DCT coefficients shows tremendous gains can be had there. After digging into jpeg2k I was never a big fan. IMHO for image compression there's a balance between algorithmic simplicity, computational efficiency and image quality. I'm looking for an article some months back about DCT coding sans huffman, but here's a quick page on how using arithmetic coding in the JPEG standard cuts jpeg file sizes down 5-12% http://www.rw-designer.com/entry/1311 http://www.rw-designer.com/entry/1311 Okay one DCT based recompressor is "lepton" which is 20% smaller than jpeg with the identical DCT quality: https://blogs.dropbox.com/tech/2016/07/lepton-image-compression-saving-22-losslessly-from-images-at-15mbs/ https://blogs.dropbox.com/tech/2016/07/lepton-image-compress...

niftich 10y ago

Other commenters are right, licensing was a big factor, because the 1990s were marred by a few high-profile events where patents on popular formats were suddenly enforced [1]. These events led to the PNG format being developed as an alternative to GIF, and Vorbis developed as an alternative to MP3. By the time JPEG 2000 rolled around, wariness of patents was well-established. On the technical side, JP2 produced unsightly ringing distortions much more noticeably at medium compression ratios than JPEG did while also blurring detail, while JPEG would only experience ringing around sharp transitions, and would get blocking artifacts elsewhere. For some people, JPEG's blocking would impact a sort of 'fake sharpness' in the image, while JP2 would produce a blurry smudge. Dark Shikari, longtime dev on x264, wrote about this phenomenon [3], which also adversely affected early H.264 encoders and made them look worse compared to ASP (like Xvid), despite former being an objectively better format. [1] https://en.wikipedia.org/wiki/GIF#Unisys_and_LZW_patent_enforcement https://en.wikipedia.org/wiki/GIF#Unisys_and_LZW_patent_enfo... [2] https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and_legislation https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and... [3] http://web.archive.org/web/20141124213159/http://x264dev.multimedia.cx/archives/317#more-317 http://web.archive.org/web/20141124213159/http://x264dev.mul...