Infrared modified 5d2 and dual ISO

[zeno2358]

I have been using ML for a few days now and am very, very happy.  One of my bodies is a 5d2 with a lifepixel enhanced color IR modification.  Everything seems to work better than fine on this body (movie mode has even begun to work again; the mod disabled it).  With this body it's necessary to do a custom white balance or the images appear all red in post.  ETTR+dual ISO seems to work fine until the images are processed with cr2hdr and the white balance of the resulting images is pushed back to red; properly blended but red.  Is there any way to tell cr2hrd to keep my custom white balance?  Or perhaps something else is going on that I need to be aware of?  Brand new at this so please be gentle.

Thanks!

[Audionut]

Well it's raw, so the WB is not baked in.  Create an import preset in LR, or something similar for other post processors, to have the software automatically apply your wanted WB.

To do it with cr2hdr automatically, you would need to use the experimental 20bit cr2hdr, and the correct WB option.

Code Select Expand

--wb=%f,%f,%f",   "use custom RGB multipliers (default 2,1,2)

You should be able to use something like UFRaw to determine the correct RGB multipliers for your modded camera.  Then you can create a shortcut to cr2hdr and modify the target parameter so that it automatically applies the RGB multipliers.

[garry23]

@zeno2358

I have a lifepixel converted 50D and have experimented with dual and ETTR.

My personal findings are that ETTR is a great asset, but dual adds little.

I have concluded, maybe wrongly, that the 'tonal space' that the IR converted camera captures, eg 720 to say 830 is easily captured in a single frame.

I would welcome hearing other experiences on using ML enhanced IR photography.

BTW here are a few IR ETTR snaps I took the other month http://photography.grayheron.net/2014/10/scale.html

[zeno2358]

Thanks for y'all's replies.  I've determined the same for now; ETTR is beneficial but several factors make dual ISO problematic at the moment.  I'll experiment with the 20bit cr2hdr and see how that goes.  (initial trials are hopeful...)

Thanks again!

[SpcCb]

zeno2358 is not wrong somewhere; --wb is a very useful option in the 20-bit cr2hdr however there's no way to simply copy the CR2 WB in the DNG (?), in case of special WB.
I use a mod camera too and do it in post, but it's not really friendly for regular daylight photos.

Maybe the CR2 WB have complex parameters, not easily transposable in the DNG (?)...

[garry23]

Like many IR shooters I 'only' convert to B&W, ie I don't try are create pseudo colour IR images.

Bottom line : I don't really get excited about my colour space; but I do rely on the ETTR.

[zeno2358]

The 5d2 has the lifepixel super color IR filter mod installed.  The color space rendered by this filter is engaging.  It looks like that if I can get the custom WB from the exposure or model the RGB response and pass it to cr2hdr that all will be good.  Though for the time being ETTR is working well all by itself and I'm amazed by ML.

Thanks for the input.  It's set me on a good track.

[garry23]

@zeno2358

I have set up a custom WB using the simple 'take an image of live grass' trick. This gives me a reasonable starting point in hand camera.

I then created a new profile for LR using the Adobe DNG profile tool.

And of course I set up a channel swop shortcut in photoshop.

But I still 'always' go B&W in the end :-)

[a1ex]

there's no way to simply copy the CR2 WB in the DNG (?), in case of special WB.

That's right - how Canon encodes WB is still a bit of a mystery. We know how to interpret the Kelvin value (e.g. with WB routines copied from ufraw), but we don't know yet how to interpret custom RGB multipliers or WBShift values.

You could try converting the CR2 to DNG with Adobe DNG converter, then copying the WB tags with exiftool. IIRC, you only need to copy AsShotNeutral.

[SpcCb]

That's right - how Canon encodes WB is still a bit of a mystery. We know how to interpret the Kelvin value (e.g. with WB routines copied from ufraw), but we don't know yet how to interpret custom RGB multipliers or WBShift values.

You could try converting the CR2 to DNG with Adobe DNG converter, then copying the WB tags with exiftool. IIRC, you only need to copy AsShotNeutral.

Indeed, Canon WB definition looks a bit complex. I thought RGB multipliers (what we can set in the ML WB menu) was enough to get a good render but I never get as good/fine results (ie. like done in the camera by Canon or done in post with a special software).
BTW we can see some of the Canon parameters in a CR2:

Code Select Expand

SensorRedLevel                  : 0
SensorBlueLevel                 : 0
WhiteBalanceRed                 : 0
WhiteBalanceBlue                : 0
ColorTemperature                : 2500
PictureStyle                    : User Def. 1
DigitalGain                     : 0
WBShiftAB                       : 0
WBShiftGM                       : 0
MeasuredRGGB                    : 948 1024 1024 963
ColorSpace                      : Adobe RGB
VRDOffset                       : 0
SensorWidth                     : 5792
SensorHeight                    : 3804
SensorLeftBorder                : 168
SensorTopBorder                 : 56
SensorRightBorder               : 5783
SensorBottomBorder              : 3799
BlackMaskLeftBorder             : 0
BlackMaskTopBorder              : 0
BlackMaskRightBorder            : 0
BlackMaskBottomBorder           : 0
ColorDataVersion                : 6 (50D/5DmkII)
WB_RGGBLevelsAsShot             : 1414 1027 1027 1598
ColorTempAsShot                 : 3898
WB_RGGBLevelsAuto               : 1876 1024 1024 1978
ColorTempAuto                   : 4122
WB_RGGBLevelsMeasured           : 1597 2354 654 1463
ColorTempMeasured               : 4122
WB_RGGBLevelsDaylight           : 2217 1024 1024 1691
ColorTempDaylight               : 5200
WB_RGGBLevelsShade              : 2539 1024 1024 1398
ColorTempShade                  : 7000
WB_RGGBLevelsCloudy             : 2383 1024 1024 1524
ColorTempCloudy                 : 6000
WB_RGGBLevelsTungsten           : 1668 1079 1079 2799
ColorTempTungsten               : 3200
WB_RGGBLevelsFluorescent        : 1950 1044 1044 2517
ColorTempFluorescent            : 3674
WB_RGGBLevelsKelvin             : 1341 1116 1116 3832
ColorTempKelvin                 : 2497
WB_RGGBLevelsFlash              : 2399 1024 1024 1509
ColorTempFlash                  : 6127
RawMeasuredRGGB                 : 0 0 0 0
CustomPictureStyleFileName      : BlackLift4

It looks there's a color matrix based on a discreet 4 pixels from the Bayer Matrix plus a magenta/green shift, modulo the Picture Style effect.
Maybe 'WB_RGGBLevelsAsShot ' seems to be the custom output matrix, what we should use to compute the RGB WB in the DNG (?) : Each params of the WB_RGGB* seems representing levels get from a reference.
For example, with the 5D2, when WB is set on Daylight the RGGB matrix is automatically defined with RGGB -> 2217 1024 1024 1691. Witch is false with a mod camera because we get more R and _obviously, but it depends of the modification_ more B. So the matrix should be adapted with RGGB something like 14?? 1024 1024 15??. We can see 1414 1027 1027 1598 for my camera, calibrated for Daylight.

Beside, I don't see why we can't change the RGB multipliers (in the ML WB menu) with +/-10E-3 increments (?). Changes made bumps of several .xx ? (sorry always away from my dev computer so can't browse the code..) Because I have precise RGB multipliers (@+/-10E-3) computed in post for my camera from very RAW files but I could not set it in the camera to see how the Canon WB_RGGB* are affected...