2012-12-23, 03:43
If I win the lotto tonight, first thing I'm buying is an Eizo. Priorities, right?!
2013-02-13, 21:24
I read the whole thread, and I'd like to know something.
Is it possible for xbmc i be icc aware in the future?
Is it possible for xbmc i be icc aware in the future?
2013-02-13, 22:14
Yes in theory. Basically a shader based transform would occur on the fly to the video. I think for a single image no issue (picture viewer)... but for a video stream? I suspect it can be done but you'd need to get a c++ coder on board. The picture viewer could probably even be done in Python as a sort of demo and I have thought about it a bit... But I suspect for a good c coder it would be a question of patching in some littlecms code into the right spots.
2013-02-14, 03:46
Just to cross post as I think people who find this thread will have an interest:
Here's a real expert talking about this stuff:
www.avforums.com/forums/video-scalers-vi...our-calibration.html
...basically, the best current way is to buy a good TV with really good calibration controls and use that. In normally available TVs, the best (by a good margin) I have found are the Panasonic plasmas (V series and up) - these are great, out of the box, and quite tunable from there. And end up being very good displays (and I am judging against those Eizos and NECs I mention....). I have yet to see any LCD (LED backed or not) - come close from the colour perspective.
I won't comment on what XBMC is doing - as I am not entirely sure. But it's definitely not doing an ICC transform. In theory this can occur in a GPU shader and thus be applied to everything coming out of XBMC. But you'd need a good C++ coder to do it. Intially I thought www.littlecms.com might help but I think this won't support GPU acceleration. Note the onyl software that really does this that I know of is MPC-HC (sourceforge.net/apps/trac/mpc-hc/wiki/New_Renderer_Settings).
I'd ask FernetMenta about this - he's got the skills, is an xbmc team member, and has a keen interest in video quality...he'd be one good man for the job I'd say, if you can get him interested.
To more directly answer your question (even if it wasn't to me!) - it depends on the quality of your display and how close it was post calibration. Most TVs still need signifcant curve adjustments in the profile and thus the answer is normally 'not very close'. My Panny V series is pretty good, though. I tried a high end Bravia recently - bloody disaster. LEDs are in general bad for colour - few colorimeters read them well (you really want an i1Display Pro for this) - and they are naturally HUGE in blue, so once you pull all this out, there's just no room left for other adjustments.
(I am at work so written in a rush, but hopefully useful!).
Where XBMC REALLY needs it is for the picture viewer (slideshow) - the eye is much more tolerant to error with video, so while nice there, it would make a really BIG difference for slideshows. I use this all the time in XBMC and would really like to see some better accuracy happening. I have been tempted to write a python addon for it, but I suspect it would be too slow to really do what I want. Better to have someone with some c++ skills do it right, in the GPU.
Also - here's basically some stuff to spell this out:
http.developer.nvidia.com/GPUGems2/gpugems2_chapter24.html
...just in case there is a shader coder with an interest lurking
See Openelec discussion here:
http://openelec.tv/forum/67-display/6150...mitstart=0
Here's a real expert talking about this stuff:
www.avforums.com/forums/video-scalers-vi...our-calibration.html
...basically, the best current way is to buy a good TV with really good calibration controls and use that. In normally available TVs, the best (by a good margin) I have found are the Panasonic plasmas (V series and up) - these are great, out of the box, and quite tunable from there. And end up being very good displays (and I am judging against those Eizos and NECs I mention....). I have yet to see any LCD (LED backed or not) - come close from the colour perspective.
I won't comment on what XBMC is doing - as I am not entirely sure. But it's definitely not doing an ICC transform. In theory this can occur in a GPU shader and thus be applied to everything coming out of XBMC. But you'd need a good C++ coder to do it. Intially I thought www.littlecms.com might help but I think this won't support GPU acceleration. Note the onyl software that really does this that I know of is MPC-HC (sourceforge.net/apps/trac/mpc-hc/wiki/New_Renderer_Settings).
I'd ask FernetMenta about this - he's got the skills, is an xbmc team member, and has a keen interest in video quality...he'd be one good man for the job I'd say, if you can get him interested.
To more directly answer your question (even if it wasn't to me!) - it depends on the quality of your display and how close it was post calibration. Most TVs still need signifcant curve adjustments in the profile and thus the answer is normally 'not very close'. My Panny V series is pretty good, though. I tried a high end Bravia recently - bloody disaster. LEDs are in general bad for colour - few colorimeters read them well (you really want an i1Display Pro for this) - and they are naturally HUGE in blue, so once you pull all this out, there's just no room left for other adjustments.
(I am at work so written in a rush, but hopefully useful!).
Where XBMC REALLY needs it is for the picture viewer (slideshow) - the eye is much more tolerant to error with video, so while nice there, it would make a really BIG difference for slideshows. I use this all the time in XBMC and would really like to see some better accuracy happening. I have been tempted to write a python addon for it, but I suspect it would be too slow to really do what I want. Better to have someone with some c++ skills do it right, in the GPU.
Also - here's basically some stuff to spell this out:
http.developer.nvidia.com/GPUGems2/gpugems2_chapter24.html
...just in case there is a shader coder with an interest lurking
See Openelec discussion here:
http://openelec.tv/forum/67-display/6150...mitstart=0
