Originally Posted by ComputerCowboy
Wow that color correction is extreme!
Yes it is. If you think of it the other way around, your jpeg picture shows you exactly how it would look through the filters with NO color correction!
Not fast enough for real-time on my i7 using separate 1080P left and right files
I am using an AMD 1090T / 6 cores / one thread per core @ 3.2GHz along with a 4Gb DDR3 RAM.
What I can do in real time without a glitch:
- 3D half SBS with color correction (one file 1920x1080)
- 3D full SBS with color correction (one file 3840x1080)
- 2D 720p with frame doubling (as per http://forums.guru3d.com/showthread.php?t=288017
- 2D 1080p with frame doubling
Not powerfull enough to do color correction AND frame doubling, even in halfSBS.
I am sure your i7 can handle this as well.
A trick I use : Force Zoom Player to use affinity with the first 5 cores (imagecfg.exe) so that ffdshow/AVIsynth can fully use the sixth core. Not sure this will work with Intel's but for my CPU its definitely an improvement.
I built a program with WPF that rips 3D Blu-ray with eac3to then constructs the right eye stream from the left and the delta and encodes it at the left eye bitrate with x264.
Wow, your are an expert. I do not pretend to be the same. My skills are limited to basic-like programming. I make extensive use of WinBatch and a little C++ but thats all.
I plan to integrate the Dobly correction as one of the output options
Remember that color correction profiles strongly depend upon the kind of projectors you use, and specially the lamps.
If you use the same typical UHP 200w lamp that is very common now, my tables should be ok for a start. This is not the case if you use a Xenon lamp for example (which is better than UHP but you will have to recalibrate).
Ideally, specific calibration should be done for each kind of projector. I intend to recalibrate once a year.
CALIBRATION. Creating the Dolby125.txt profile.
I have built a set of Winbatch (iterpreter) routines to show/edit calibration patterns on the screen.
Unfortunately, these are not "polished", but still usable by anybody. The best would be to have winbatch installed on your system, so you can edit the routines at your wish. Otherwise, I can compile the routines to EXEs for you.
I am writing a help file on how to use the routines and make the calibration. That may take a couple of days.
Basically, you have to set your graphic output to 2x1080p (3840x1080) and let the graphic card driver split the screen in half (affinity or horizontal span if ATI, whatever? if NVIDIA). All my patterns are 3840x1080.
Originally Posted by darinp2
In case it helps anybody, the first thumbnail for this post:http://www.avsforum.com/avs-vb/showt...6#post19426946
has the spectrums that each lens in the Dolby glasses lets through. I was thinking that the Sony VW200 might work well for a dual projector system with this system given the Xenon bulb, but it looks like even its spectrum is far from ideal.
Very interesting. That proves you have to calibrate for each projector. Infitec claims that they can do color correction for almost any kind of projector, including the UHP type, which is the worst case, as show at the end of http://www.jumbovision.com.au/files/...hite_Paper.pdf
(terrible, isn't it?). UHP lamp is cruelly lacking red....
By the way, remember that Infitec and Dolby DO NOT use the same wavelengths. So stick with only one supplier. My sample profiles/tables are for Dolby filters/glasses.
Still I completly disagree on:
Therefore it might be difficult to just put these glasses in front of a normal projector without loosing too much light. This since you would need to tame all the other five colors to match the lowest one.
This is true only if you keep the primaries as such and only adjust their levels. But this is not the case when correcting. You have to matrix the colors (thats the reason why you cannot do that with the card driver or the projector settings).
For example, for the left filter, pure red [255,0,0] is converted to [255,68,0]. It means you add some green, still keeping red red. And because green is not blocked at all by the left filter, yo add a huge amount of brightness.
All along the calibrating process, you will be facing a tradeoff between color accuracy and overall brightness. I found that a 40% transmission factor (60% light loss) gives an enough accurate color spectrum, even in reds.