Summary: | FILEIO : add new option to convert automatically JPEG file in 16 bits color depth | ||
---|---|---|---|
Product: | [Applications] digikam | Reporter: | Christian <buitk14> |
Component: | Plugin-DImg-JPEG | Assignee: | Digikam Developers <digikam-bugs-null> |
Status: | REPORTED --- | ||
Severity: | wishlist | CC: | caulier.gilles, cfeck |
Priority: | NOR | ||
Version: | 4.1.0 | ||
Target Milestone: | --- | ||
Platform: | unspecified | ||
OS: | All | ||
Latest Commit: | Version Fixed In: |
Description
Christian
2014-07-22 11:54:04 UTC
Can you provide a link to the specification of 16 bit JPEG? (In reply to Christoph Feck from comment #1) > Can you provide a link to the specification of 16 bit JPEG? Hi Chris, The latest Jpeg specs are all restricted to 8 bit. The conversion to 16 bit color depth in digiKam should take place after opening an 8 bit jpg without any user action to minimize computation errors. The conversion back to 8 bit is already implemented. Christian Ah, I now understood this request. You want the JPEG loader to convert the data to 16 bit internally before doing any further operations, so that you do not have to do it manually after each load. Makes sense. Converting 8 bits image to 16 bits will introduce noise to fill histogram holes : https://projects.kde.org/projects/extragear/graphics/digikam/repository/revisions/master/entry/libs/dimg/dimg.cpp#L2757 Gilles Caulier (In reply to Gilles Caulier from comment #4) > Converting 8 bits image to 16 bits will introduce noise to fill histogram > holes : > > https://projects.kde.org/projects/extragear/graphics/digikam/repository/ > revisions/master/entry/libs/dimg/dimg.cpp#L2757 > > Gilles Caulier I am not an expert in digital imaging. But as far as I can see the results, the artificial noise is a minor problem compared to artefacts caused by 8 bit rounding errors. The noise is added to the 8 bit version before the transformation takes place? Or in 16 bit mode? Noise in 8 bit mode would be a drawback for jpgs with less than 12 MP. Well distributed entropy in 16 bit mode might be cleaned when the colors are transferred back to 8 bit mode. I can't tell. My results where fine in print and on a high resolution / wide gammut monitor using argb profile for 20 MP jpgs. But this is no proof in any way. I will continue using the 16 bit conversion before I apply gradiation and the like - artefacts/gradients are worse than noise for me. Look in code, noise is apply to 16 bits image during conversion from 8 to 16. Gilles Caulier (In reply to Gilles Caulier from comment #6) > Look in code, noise is apply to 16 bits image during conversion from 8 to 16. > > Gilles Caulier Thank you. Auto-Conversion to 16 bit would be a great feature for me. Dont know if is wanted by others. Maik, Technically, this one is very simple to implement, as all background code is already here. A new option must be add to FileIO settings to convert automatically JPEG to 16 bits color depth. This must be available only for Editor and BQM. Other part of digiKam must still use 8 bits color depth for performance reason. In JPEGLoader, loading file must apply DImg::convert8to16bits method in case of option is on. Gilles |