“Beliefs are ideas going bald” a remark credited to Andre Breton et al.

Details regarding JPEG 2000

  • A GPU decoder: "For anyone who has a need for speed, check out my OpenCL jpeg 2000 codec. I am currently able to decode RGB lossless images using the GPU: https://github.com/OpenCodec/ThousandthChicken
  • A jpeg 2000 decoder written in javascript: https://github.com/mozilla/pdf.js/blob/master/src/core/jpx.js
  • openjpeg notes for year 2015 "0) Moved to GitHub: git work flow, pull requests instead of patches, integration with Appveyor and Travis 1) Fixed a huge number of bugs. I would guess that the library is more stable and correct than it has ever been in the past.2) Improved the test suite: more test cases, integration with JPylyzer 3) Became an official reference implementation of the Jpeg 2000 standard"

Lossless mode: Experiment from Forum. The compression is: 40421376 >> 15735885 (0.389 ratio) (in 10s) rearrange raw image to interleave all bayer components, and used -F 2752,1836,4,12,u , but the compression is worse: 19175592 (ratio 0.474) with Bzip2, the result size is 17934600 (0.444 ratio) (in 8.3s)

"Lossless j2k is not magic. PNG will out-perform it with normal image sizes. Where j2k comes into its own is with massive images, like GIS datasets. j2k can efficiently decode a small section of the image, which PNG and such is not designed for." "Why not de-bayer it? Do you really expect to use the image without de-bayering it? Use the highest quality de-bayer you can find and be done with it."

  • OpenJPEG 2.1.0 has been released (April 2014):

https://sourceforge.net/projects/openjpeg.mirror/files/2.1.0/ (*)

"Compared to the 2.0.0 version, a huge amount of bugs has been fixed, the code has been cleaned, and the library is now much more stable than before. You’ll find more info about this release here: http://openjpeg.googlecode.com/svn/tags/version.2.1/NEWS and a detailed view of all the changes that have been made here: http://openjpeg.googlecode.com/svn/tags/version.2.1/CHANGES

Despite a few needed changes in the API, we strongly encourage you to switch to this new version to benefit from the new features and bug fixes. Moreover, this new release is in the process of becoming an official reference JPEG 2000 software of the JPEG committee (www.jpeg.org), and as such, it will be part of the JPEG 2000 standard (Part-5 : reference software). Official approval will be in October 2014. As you may have noticed, we have also released maintenance versions of OpenJPEG : 1.5.2 and 2.0.1. Please note that the 1.5.x and 2.0.x branches will not be improved in the future and will only receive critical bug fixes, if any. A lot of work is still to be done to make OpenJPEG faster and more robust so don’t hesitate to contribute and submit your comments and ideas (here : https://code.google.com/p/openjpeg/issues/list)."

  • From the open source sw point of view the OpenJPEG is releasing 2.1 now. This is planned to be an official JPEG2000 reference code to be ratified over the Summer.


They have mainly been fixing bugs etc but would be a good test reference place as well.



(Historical reasons they use both google code and source forge repositories)

- Original staff net item:

"Andrew McGregor (RAL Space) Awarded 94,656 to develop a JPEG2000 image compression core (VHDL software) that can be ported into a radiation-tolerant Field Programmable Gate Array (FPGA) This will be the first radiation hard, non-ITAR controlled, JPEG2000 compression core available on the market and will complement and process the data generated by our camera electronic boxes. This will allow RAL Space to offer customers a complete camera system. The camera electronic boxes currently provided by RAL Space sell for circa 500,000 each. In contrast a camera system with integrated image compression could sell for circa 3 million; a six-fold gain in system 'value-added' and return to STFC. Hence, this new JPEG 2000 compression system will add further to the system-level product chain, and allow RAL Space to bid for significantly increased end-price programmes."

Note and Thoughts:

Article: http://petapixel.com/2015/09/12/jpeg-2000-the-better-alternative-to-jpeg-that-never-made-it-big/

Article: http://nevion.com/news/press-releases/2015/nasa-successfully-verifies-interoperability-vsfs-tr-01-jpeg-2000-video-over

Clustertech: have image compression in FPGAs http://ac.clustertech.com/our-product-2/our-product/

Request for non-RGB input: OpenJPEG can process YCbCr data: you have to combine the -F option (with image dimensions, signedness and optional subsampling) and -mct option (-mct 0 in this case to tell OpenJPEG to disable RGB->YCbCr conversion as you already did it). You might have to change the extension of your input file to raw Type opj_compress -h for detailed syntax

J2K work related to FPGA: "latest work on the cat board. Lattice ICE-40 cat board The forward & inverse dwt is working [1] https://github.com/develone/jpeg-2000-test/tree/master/pc_fast_blinker_jpeg

GPU solution in a commercial system: http://www.tvtechnologyeurope.com/post-production/vizrt-viz-engine-gets-jpeg2000-from-comprimato/01077

Other related software: Jpylyzer http://jpylyzer.openpreservation.org/2015/12/03/Release-of-jpylyzer-1-15-1/


Is J2K slow: comment at: http://blog.hexagongeospatial.com/jpeg2000-quirks/?utm_medium=twitter&utm_source=twitterfeed#prettyPhoto


Model of the human visual system for the evaluation of perceptual image quality of compressed images K Roubik, V Slaninka, J Dusek, M Klima, M Bernas (Department of Radiolectronics, CTU FEE Prague, Technicka 2, CZ 16627 Prague 6, Czech Republic; e-mail: roubik@feld.cvut.cz)

-- MartinTurner - 03 Apr 2014

Edit | Attach | Watch | Print version | History: r15 < r14 < r13 < r12 < r11 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r14 - 29 Dec 2015 - MartinTurner
  • Edit
  • Attach
This site is powered by the TWiki collaboration platform Powered by Perl This site is powered by the TWiki collaboration platformCopyright & by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback