Jump to content
UltravioletPhotography

In-camera White Balance


rfcurry

Recommended Posts

I am wondering how many here use the in-camera custom WB against the framed area rather than a 99% WB standard. I enjoy using the background itself much of the time, as it gives a more varied coloring. Below are some examples:

 

Photos taken with Lumix GF1 modified to full spectrum, a CZJ Tessar 50/2.8 lens, f8, ISO 400, natural sunlight with no clouds, and a UV bandpass filter.

 

First, visible light @ ISO 100, f8, 1/250s. PP one-click WB and reduce to 1000px width.

post-19-0-24068900-1475776864.jpg

 

Background WB. PP one-click WB and reduce to 1000px width.

post-19-0-91044600-1475777031.jpg

 

PTFE disc WB. PP one-click WB and reduce to 1000px width.

post-19-0-70786200-1475776959.jpg

 

 

Using a different UV bandpass filter but otherwise the same.

 

Background WB. PP one-click WB and reduce to 1000px width.

post-19-0-04003900-1475777146.jpg

 

PTFE disc WB. PP one-click WB and reduce to 1000px width.

post-19-0-64378700-1475777198.jpg

 

Do many use the background for setting in-camera WB?

Thanks.

Link to comment
If I understand the capture process correctly, in-camera white balance is only applied to image data which is converted in-camera to a jpg. Image data saved as RAW is basically 'undeveloped', thus has no white balance, or other, development settings applied. Since I shoot almost exclusively in RAW any/all settings are applied post-process, including white balance. In this mode I take/set white-balance from some area within the image. I understand this does not technically provide for comparability of images shot under varying illumination sources, but since anything beyond the limits of human vision is technically 'colorless' such images are false color anyway - so I don't look to achieve any kind of color correctness, only luminance correctness (for what its worth). Also, I have a very technical job, so UV photography is a qualitative hobby for me (though I do enjoy reading some of the more technical/detailed info on the forums here).
Link to comment

Most converters these days can "read" the white balance used in your camera and provide a conversion which is the same as or close to what you set in the camera. The beauty of having a good in-camera white balance for UV is that it makes it easier to focus and shoot UV using Live View. But if you use the camera's auto white balance or daylight WB, for example, then you get stuck with a lot of saturated reds and magentas which are very hard to work with while shooting. Most converted mirrorless cameras cameras can set WB through the UV filtration to save in a settings bank. DSLRs cannot always set WB through the UV filters. I had some success with my converted Pentax K5 DSLR at setting an in-camera WB through some UV filters but it was not always easy. The Nikon converted DSLRs will not do this at all. But there is a way to "push" the white balance setting in a Nikon DSLR to reduce the red/magenta by a large amount. So I use that and I'm then able to see through the UV filtration when in Live View. I convert about 50% of my Nikon UV photos in one of the Nikon converters so that all in-camera settings are preserved. For the other 50%, I convert in Photo Ninja which has a superb "click-white" balance tool and the capability to create and save white balance presets. So it all works out in the end however you approach it.

 

Well, that was rambling......... :D

 

I'll try to get the method written up for pushing the red/magenta out of a Nikon DSLR white balance preset. It is iterative and takes some patience to achieve. But it can then be saved in one of the white balance banks for easy use while shooting. The offending reds & magentas become pale oranges, pinks and some yellows.

 

*****

 

In answer to Reed's question above: I often set in-camera WB "against the scene" or against the ground when shooting Infrared with the GH1. I like the results I get with that in IR. For UV I always shoot a picture of the white standards or the white PTFE disk. I think I got into that habit for UV work because Bjørn and I wanted the standardized blue/yellow/gray appearance for the botanical shots we both collect so we would have a uniform appearance across different platforms for the same species. Also because I use the Nikons for that work (which needs lots of pixels and good IQ) and, as mentioned, the Niks can't do UV WB in-camera.

Link to comment

Reed, Your results with the second and even third shot are quite interesting. The second shot is very nice.

 

Setting the WB with the background/ground/whole frame, is this anything similar to a full frame CNX2 marqee, or even a large background maquee?

I often use the background for the white balance in CNX2, and it works nice. I tend to have odd results using PTFE, and a background WB or full frame WB usually look the best to me.

 

Andrea, with my Nikon, what I did was shot PTFE for a preset WB, then I shot the PFTE again using the first PTFE WB preset, making a second preset WB from that, then I shot the PTFE again using the second preset WB, making a third preset WB from that. Tricky, huh? ;-) Well, it gets whiter and whiter with each generation. It still isn't that good, but better than any other in camera WB I can make.

Not very orthodox, but it renders something that is easier to work with in live view, and a little closer to 'norBal', but I always white balance everything later from the RAW file.

Link to comment

Steve,

 

I like the #2 shot, as well. Obviously, that particular filter is passing different UV wavelengths from the Rudbeckia than we usually see. We are accustomed to a UV dark - low reflectance - target on the Brown-eyed Susan. Those yellow images suggest that we have yet to have a full understanding of the flower's target. I wonder what the wavelengths are that we are seeing here?

 

Thanks for your input.

Link to comment

Reed, if you extract the raw colours using an app like Raw Digger, then that will often give you some information about how the filters are working. If you like, put the raws in Dropbox and I will run them through RD for you.

This two-tone yellow appearance of the Rudbeckia is what we would expect when using a filter which passes violet/blue/blue-green to create a "bee vision" photo. Yellow areas which absorb UV appear yellow/yellow-green/green to the bee (insofar as we can extrapolate from bee visual receptor structure). Yellow areas which reflect UV are a special colour to the bee which we cannot imagine.

 

Steve, that iterative process for in-camera WB uses the same idea I use to create my low-red white balance preset for my D600.

Link to comment

Andrea,

 

I will try to squeeze one more image from my poor, tattered Rudbeckia :( and then put the RAW in dropbox.

Thanks.

Link to comment

Reed sent me two raw files to run through Raw Digger for an attempt to deconstruct what was recorded. It is all very interesting. And I am eager to have comments about what I'm about to present. Have I made valid observations and conclusions?

 

Here are the JPG extractions from the raw files. These photos were made with a UV-pass filter which sees the Rudbeckia flower's UV absorbing area as False Yellow after white balance is applied.

The 500 px wide photos will display side-by-side if your browser is expanded enough.

A_P114013802.jpgB_P114014301.jpg

 

 

Here are the extractions of the raw colours via Raw Digger. (This is called the Raw Composite in RD). NO white balance step was applied, so this represents only the demosaiced colours to which intensity information has been applied (saturation and brightness).

  • These JPGs do have an sRGB profile added for necessary colour management but that does not affect WB, per se. Adding ICC profiles simply tells the monitor and the browser how to translate the file colours into their own colour space.

  • Color dropper samples were made on various areas in each photo and stamped onto the photo in PSE. I was attempting to find the hue range with these samples.

  • Sampled colours were then pushed to their purest 100% saturation and 100% brightness and stamped onto the photo.

  • The labels indicates the degree of Hue for the sampled colours in the 360° colour circle. Between the two photos there is a 1-2 degree difference in Hues. This is negligible. It could be due to sampling error or due to the fact that outdoor light is never constant in composition or intensity. That Hue degree is, of course, the same for both the less saturated, less bright sampled colour and its pushed version in the HSB model.

  • Across the top of the photo is a gradient of the Hue range.

These photos are 1000 px wide.

A_P1140138rawCompCol.jpg

 

B_P1140143rawCompCol.jpg

 

 

 

Now I'll show you the raw channels from the A photo. (The B photo would be similar, so no need to present both.)

Each photo is 300 px wide, so you should get a side-by-side display.

A_P1140138_redChan.jpgA_P1140138_greenChan.jpgA_P1140138_blueChan.jpg

 

 

This is the raw composite of the A photo with 100% saturation applied in PSE using the Hue/Sat tool.

A_P1140138rawCompFullSat.jpg

 

 

CONCLUSION:

Apologies for letting the readers down, but I'm a little short on conclusions at this point. :D

What can we conclude?

The UV recording was made mostly in the red channel??

Link to comment

Andrea,

 

A million thanks for all the hard work!

 

We can conclude that the UV-dark target of Rudbeckia is not dark in all UVA wavelengths. Obviously, we are passing different UV wavelengths in these images, using this particular filter, than we usually pass. We know that there is no NIR contamination -- I included a 30s, Hoya 72, f8, ISO 400 RAW file in the dropbox, just to make sure. That UV-bandpass filter records OD3 at 397nm and no discernible transmission to 1100nm (caveat: my Hitachi U-1500 spectrophotometer doesn't go lower than OD3). So, I am fairly certain all that is present in that image is reflected UV. But, what wavelengths? Does anyone have a method of extrapolation from the data that Andrea so graciously provided?

 

Here is the coloration of the UV-dark target as we usually see it (represented by 4th square down on right):

post-19-0-64457600-1475970086.jpg

 

and here is the coloration with only the brightness altered, increased from 22% to 57% (represented by 3rd square down on right)

post-19-0-29544600-1475971881.jpg

 

Finally, here is the color panel Klaus made some time ago. The 370nm (rgb 142 138 109) is almost spot on with the brightened UV-dark area (143 140 108):

post-19-0-25871500-1475972089.jpg

 

Coincidence, probably. It is extremely unlikely that only a single UV wavelength is recorded by the sensor for that pixel. Ah, well, if anyone finds a correlation between coloration as interpreted by the sensor, and actual wavelengths, there is hope for a Nobel prize in your future... or even an Ignobel prize.

Link to comment

The "red" channel in most cameras is better at capturing the wavelengths below 370nm. This has been discussed elsewhere.

 

Where image chrominance is low, hue error tends to be large, and more susceptible to disruption by casts and cast gradients. I notice that several of the sampled areas appear to have low chrominance. Keep this in mind. This effect can make it appear that there is more chromaticity in the image than is actually the case.

Link to comment

Reed, no, it is impossible to go backwards from sensor colour to wavelength because are multiple wavelength combinations which can produce the same colour.

 

Just recently I looked at the raw colours in two photos made with a very narrowband 340 nm peak filter and found hues from approximately 30° - 50° (depending on illumination) which were certainly not any of the greens shown in the color map above. One of my 340/10 photos did produce greens after white balancing. The other one did not. There are so very very many variables which affect the false colours we get in our UV photos. The point here is that this colour map was made a certain way and we don't know how. And I can't match those map colours in a lot of my UV work.

 

I'm also not really sure if it is valid to be looking at the raw colours in a UV photo and drawing any conclusions from them? I started doing this because I just wanted to see in which channels the UV light was landing when I made a UV photograph. Or to see what the width of the hue range was. On the other hand, is it valid to look at white balanced false colours in a UV photo and draw any conclusions from those colours? Which set of colours would I use to make a colour map??

 

I get bogged down in this stuff. Which means it is time to simply go make photographs and quit worrying about it. :lol:

 

No greens here in the hues taken from the raw versions of my 340/10 photos.

340hues.jpg

 

*************

 

Reed, we must also rule out the effects of visible contamination from the filter you used to make the Rudbeckia photos.

Link to comment

I would go with a different approach:

 

What do you intent, when you use white balancing?

Simple speaking: You want something look white, when it is white.

 

When your light source is close to a black radiator, then you might adjust your camera settings (WB Black Radiation) so, that your picture shows white as white.

 

Now your light source has a tint, e.g. it has too much green.

 

Now you can do two things:

 

1.) You want your pictures showing white as white. So you do a new white balancing (WB Tinted Light). Then everything is fine.

 

2.) You want to judge the bad light source. So instead you use the first white balance (WB Black Radiation). Now the white will look greenish.

 

 

 

Now you do an extreme trial. You get a new light source, with a narrow band red spectrum (a laser, or using a narrow band filter).

 

Again you can do two things:

 

1.) You want your picture showing white as white and you do a white balancing (WB Monochrome Red) with that setting. The outcome will be a black and white picture!

 

2.) You want to judge the light source comparing to your reference of a black radiator, so you use the WB Black Radiation. Now everything in the picture will look red as your light source.

 

My opinion is, all the same is true with UV pictures. If you want to measure something (by the color), you must first set up a reference light source, which covers the whole spectrum (“UV- Black radiation”) and set up a WB UV Black Radiation. Reference could be the sun (… as a first approach or a flash).

 

With slightly different light sources (“tinted UV light” or filtered UV light) covering the whole spectrum, you can do individual white balancing, so pictures will look similar to the one with the reference light (and WB UV Black Radiation).

 

When you cut the UV spectrum with a narrow band filter, again you get more a black and white picture with an individual white balancing. when you want to judge the white by colours, you need to use the reference WB.

Link to comment

A possible set up?

 

Camera + Lens + Baader U

Flash light

WB reference -target

Making a picture and doing a WB

 

Then, to test your narrow band filter, put this additionally in front of the Baader and take a picture using the WB from the previous setting.

Now your 340nm filtered target looks green?

Link to comment
To get something looking 'white', we need some response from all three channels of the camera. In case of the very narrow-band UV filters, often only one or two channels contribute. Thus, the 'white' will be green or blue or something else.
Link to comment

Alaun, good ideas. I will try to run the test on the 340 that you suggest.

 

My definition of UV-white is taken from either white PTFE or white Spectralon, both substances which reflect UV (and R, G, B, IR) with very little attenuation. (Not sure of the UV range endpoints, but they are deep.) Thus UV-grey is a less intense reflection of UV. Continuing like this, we have that UV-black is a complete absorbance of UV by the PTFE/Spectralon. My definition of UV-white takes into account also reflection of all or R, G and B by those substances. This is because we are trying to push a UV photograph into an RGB representation. So really my definition of UV-white is a simple extension of the definition of Visible-white.

 

In case of the very narrow-band UV filters, often only one or two channels contribute. Thus, the 'white' will be green or blue or something else.

This is not really happening because we have to take intensity into account. My photos from the 340fwhm10 always have data in all 3 channels.

Link to comment
I balance the 340 nm narrow-band captures against the PTFE target and the outcome is green, as expected, not 'white'. Same result whether I pre-balance in camera (Panasonic) or do a click-white against the PTFE later in Photo ninja (for my Nikons). No blue is present thus green substitutes for 'white'.
Link to comment

Someone asked me about that 2° difference above. Not about how it happens (obvious, light changes), but whether there was any correlation to wavelengths. Using the available online converters we can get some rough estimates of rgb-to-hue-to-wavelength. It looks like 1° difference is about a 1 nm difference.

 

BUT you are making a LOT of assumptions in that correlation. Do NOT go too far with this until you sort out ALL the factors involved in producing that colour. After all, I've sampled RGB off a JPG/sRGB translation of the data in Photoshop Elements. That could not be called accurate. :D

 

What you really would need to look at is the actual binary data. That can prolly be done with Dcraw but I don't really have the time at the moment to pursue this. Again, how the data is acquired is a function of so very many inputs. The sensor maps attempted by various folks have not -- to my knowledge -- looked at the really raw data, have not always had the proper spectroscopy setup, and have not explained how to properly account for white balance -- or not. And finally, but most importantly, RGB to wavelength is not a 1-to-1 mapping so how do you get around that???

 

I could go on, but will shut up now.

 

But, just for fun, I'll look at the orange hues from the preceding photo around 23 - 25°. Fully saturated versions of these hues are used in the first calculator. So either one or two of the RGB values will always be 0.

Link: Academo Wavelength to Color Relationship

Link: Rapid Tables RGB to HSL Color Conversion

 

255,115,0 621 nm 27°

255,111,0 622 nm 26°

255,107,0 623 nm 25°

255,103,0 624 nm 24°

255,99,0 625 nm 23°

255,95,0 626 nm 22°

255,91,0 627 nm 21°

 

Here is a 1000 pixel wide strip of the 100% saturated, 100% bright hues from 21° - 27°. Not a lot of change in that 6° to my human eyes. B)

21to27.jpg

Link to comment

I balance the 340 nm narrow-band captures against the PTFE target and the outcome is green, as expected, not 'white'. Same result whether I pre-balance in camera (Panasonic) or do a click-white against the PTFE later in Photo ninja (for my Nikons). No blue is present thus green substitutes for 'white'.

 

I will also run this experiment to provide a confirmation. I have not yet done so with my 340/10 because of that blasted reflection problem. Don't know what to do about that except shoot in the middle (as illustrated some previous threads). I may have to abandon this particular 340 filter. (sigh.)

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...