Talk:Lossless JPEG

From Wikipedia, the free encyclopedia

[edit] Patents?

Is JPEG-LS subject to patents in any developed country that require payment by implementors? --Damian Yerrick (☎) 14:07, 16 July 2006 (UTC)

I'm fairly sure that JPEG-LS is subject to patents, but that its license requires no payments for use (see HP's license for JPEG-LS, which I haven't read all of, to be sure). Calbaer 21:55, 17 July 2006 (UTC)

[edit] title

since this covers all 3 lossless JPEG formats shouldn't it be titled appropriately, i propose "lossless variants of JPEG" Plugwash 17:08, 31 July 2006 (UTC)

[edit] Compression Ratios

JPEG-LS does not get higher compression ratios than JPEG2000 with any of the public domain software. For a test, I compared JPEG2000 (Meesoft Image Analyzer) to JPEG-LS (University of British Columbia's implementation) on Kodak test images. (http://r0k.us/graphics/kodak/) JPEG2000 got more than a 2:1 compression ratio on all of them, and they correctly decompress losslessly. JPEG-LS did much worse; some of them were larger than the original PNG files.

If JPEG-LS is able to get higher compression ratios than lossless JPEG2000, why does every JPEG-LS implementation I try do a little better than PNG, but much WORSE than lossless JPEG2000? Korejwa 19:38, 15 August 2006 (UTC)

There are many potential explanations for this, including a suboptimal encoder, accidental use of lossy mode, accidental use of distorted images, scale of the images lending themselves less to lossless coding, or the fact that only natural images were used (where the two JPEG methods are reported to be competitive) rather than artificial or compound images (where JPEG-LS wins hands-down). A more comprehensive citation is needed for this, but JPEG's own comparison of 2000 with the older LS (UBC implementation) on seven images from the JPEG 2000 test set showed LS winning most of the time, so that's what I went by. Unfortunately, the test set is not freely available due to copyright issues. The sentences could be changed to:
  • JPEG 2000 includes a lossless mode based on a special integer wavelet filter (biorthogonal 3/5). JPEG 2000's lossless mode runs more slowly and often has worse compression ratios than JPEG-LS, especially on artificial and compound images[1]. However, it is scalable and progressive, and, because its algorithm is more similar to JPEG 2000, it is more widely supported[citation needed].

In summary, though, I don't know what accounts for the difference. Hopefully you can hazard a guess. Calbaer 21:20, 15 August 2006 (UTC)

Let's be bold and change it Joachim.Kluge 07:48, 7 February 2007 (UTC)

In my experience, JPEG-LS is better than JPEG 2000 lossless mode for grayscale images, but it isn’t the case for color images. JPEG-LS usually compress color components independently, while JPEG 2000 can employ reversible color transform (RCT) to remove inter-color correlations. JPEG-LS part 2 specifies a new color transform technique but it has not been popular. 124.87.182.116 19:33, 10 August 2007 (UTC)