WhiBal and LightZone: Am I being too picky?

[amacb]

Mike,

I've been having a discussion with the folks over at LightZone with regards to their implementation of White Balance.

Simply put, if I select a white balance tool and click on a WhiBal in an image, I expect it to render a neutral balanced image.

That's not what I get from the LightZone white balance tool. It makes some corrections, but there are still casts.

This PDF file ( http://www.macbrien.com/onlinestorage/ColorBalanceIssues.pdf) shows the differences in results from 5 different RAW processors. 4 of the 5 provided reasonably acceptable initial results from one click on either the neutral or white portions of the WhiBal. LightZone did not. Slide 12 shows them all, side by side I shared this document with LightZone and they said they'd look into it.

I finally figured out how to get a neutral image using a combination of LightZone's "white balance" and color balance tools. But this two step process (see: http://www.macbrien.com/onlinestorage/LightZoneWhiteBalance-PartII.pdf) seems both a bit clunky, and still indicative that the LZ "white balance" is not the same as most other tools. I sent this document on to LZ tech support but the issue remains as of LZ release 2.05.

My question to you and the group: Am I being too picky on this subject, or is it reasonable to expect a white balance tool to work in a standard manner?

Cheers,

amacb

[pegelli]

Dear amacb,

Don't think you're picky. There a definite greenish cast on my monitor for LightZone while the others seem neutral. I'm not a specialist and also have never used LightZone but had the following "brainwave":
Since the histogram still shows one peak more or less at the same spot in the histogram for all 3 RGB channels I assume the program considers this neutral grey. So one explanation might be that there is some kind of "stealth" ICC profile started together when you start your LightZone application. However looking at the color of the UI it's only working on the picture, and not on your whole screen, so if this is the case it's most possible some kind of print ICC.

Hope this helps solving your problem. If not it might spur others to give more expert opinions. Good luck.

pegelli

[ThomasH]

I am curious now, what LightZone said presisely about their differing "white balance" adjustment? Is it a bug or a "feature"?

[amacb]

When I initially contacted LightZone, I provided them with the first of the PDF files I linked in my original post above. They said that they would look into the issue. I followed up with the observation that using the LightZone "white balance"/color balance combo got me the result I was looking for. At that point, LightZone tech support said that they considered the issue closed. I followed up with another message to tech support letting them know the issue still remained and provided them the second PDF at that time. Since then (24 Nov) I've heard nothing more from LightZone.

[MichaelT]

You are not too picky. Also note that the Gray of the WhiBal will always be more accurate than the White. A proper WB, is achieved by allowing the RC to "know" the color of the light (as read by the very neutral WhiBal) and apply it properly during the debayering process, while at the same time applying the data from the camera profile. There are other ways to do it (ACR is a little different), but the result should be the same. And the simple test is this. After using the "White balance" tool (whatever it is called in a given RC), one can simply put the cursor over the Gray part of the WhiBal (where the tool was clicked) and the values read should be R=G=B (example R=195, G=195, B=195) give or take 1 value point. This proves the true neutrality of the image, and that is the reason for a very accurate WB reference like WhiBal.

I will do some experimentation and contact Light Crafts. Thanks for pointing this out. I consider this improper behavior in Lightzone if my tests confirm.

Happy new year.

[Fabio_Riccardi]

Hi,

thanks for the spotting, it is a bug, it only happens when there is an additional color cast to the image, i.e. for fluorescent light sources.

The WB tool has a "Tint" slider that should adjust automatically but for some reason doesn't.

We'll fix this in the next release.

Thanks again for your interest, best regards and happy new year,

- Fabio

[MichaelT]

Thanks to all. I have also found that the WB tool appears to take a single sample pixel, rather than a 3x3 or 5x5 sample. I will suggest that to Fabio.

[Fabio_Riccardi]

LightZone actually samples a 3x3 pixel area.

- Fabio

[MichaelT]

Ah...I was not sampling at 100%, and that is what led me astray. Thanks for the clarification..

[Visitor101]

I have found that version 2.0.6 is now released and in the list of fixes is the white balance fix.

I have not tried the WhiBal myself with LZ so cannot comment about this ;)

[amacb]

Downloaded the update last night. Working on the evaluation tonight and hope to have the results posted by Friday. I'm a little concerned that the fix was described as only applying to fluorescent light sources. I'll be testing with both "daylight" fluorescent and incandescent sources and comparing results to Lightroom and RawShooter Premium.

Test setup will be as follows:

- 2 light sources of identical type in 10" Smith Victor reflectors on lightstands.

- Sources placed on either side of WhiBal card and shining down at approx 45 degree angle

- Nikon D200 with Tamron 90mm f2.8 macro lens mounted on tripod and positioned perpendicular to WhiBal and focused so the neutral gray portion of the card completely fills the frame.

- Camera white balance will be set at auto

- Exposure mode will be set to manual with exposure determined by Sekonic L-508 in incident mode

Test methodology:

- Using the setup described above, exposures of the WhiBal card will be made under both types of lighting.

- The resulting .NEF files will be processed in:
-- Lightzone
-- Lightroom
-- RawShooter Premium

- In each program the file will be:
-- Imported/opened
-- Have its white balance corrected using each tool's white balance eyedropper tool to select an area in the center of the image
-- Exported to a TIFF with a filename indicating the RC used.

- Each TIFF will be opened in Photoshop CS2 and the RGB values found at the center of the image will be recorded

- Results will be posted here along with a test report in PDF.

Why, no, as a matter of fact, I DON'T have a life. Why do you ask?

Cheers,

amacb aka riverphoto

[MichaelT]

I am not pleased with the results of the fix. I will contact Fabio.

[Fabio_Riccardi]

Hi Michael,

I would love to have some more specific feedback, problem images would also help very much.

Thanks,

- Fabio

[MichaelT]

I will send,..

[amacb]

Hey Guys,

Real life intruded a little more than I expected this past week (Go Pats!), but here's the results of my testing as I described in my last post above.

The LZ "white balance" tool still has issues.

I was a little concerned when the release notes for LZ 2.06 stated: "Fixed the White Balance tool for photos taken under fluorescent light."

Um. Yeah. I think the problem was with more than just fluorescent light. My testing bears that out. You can read the report posted at:
http://macbrienwp04.home.comcast.net/LZ/LightZoneWhiteBalance-PartIII.pdf

As you'll see, the results under fluorescent light were still off a bit. Incandescent was way off.

(Fabio: Please let me know if you'd like copies of my raw files and the resultant TIFFs. I'd be more than happy to provide them. Same goes for you, Mike, if yer interested.)

Adios,

amacb

[MichaelT]

amacb,

I concur with your findings. I believe there is still a major White Balance issue with LightZone 2.06.

[Visitor101]

I think the LZ "crew" are at MacWorld, so all being well following their return there will be some positive action to resolve this issue in the next revision ~ due too soon?

:)

[Fabio_Riccardi]

Dear all,

thank you very much for helping "debugging" our White Balance Tool, in the forthcoming version 2.1 we have a new algorithm that works better under a wider array of use cases.

I'd love to have your test images to make sure thyat we actually did it right this time. Please send mail to fabio at lightcrafts dot com, the best is to use a web site where we can download from, otherwise we have a yousendit account for large files.

Cheers,

- Fabio

[MichaelT]

I have sent a good test file.

[Visitor101]

Hi All

LC have posted v2.1 here http://www.lightcrafts.com/products/lightzone/download/ no mention of WhiBal specifically so possibly they are still working on that aspect?

I look forward to those amacb & Michael confirming whether it has been 'fixed' in this release.

:)

[MichaelT]

I just tested and it is not fixed in v2.1.

[amacb]

I concur with the esteemed Mr. Tapes. My results can be viewed at:

http://macbrienwp04.home.comcast.net/LZ/LightZoneWhiteBalance-PartIV.pdf

Same files as obtained by the test setup in LightZoneWhiteBalance-PartIII (link in previous post).

Same measurement methodology.

Same "white balance" problem.

amacb