An Easier UniWB for EOS Digital Cameras (at least the XT)

Yes, it seems to work fine! So... one last stupid question... Now that I've set the WB this way I can erase the black picture from the card? Even reformat the card? And change the WB setting to something else and then back to this and the setting will still be there? Is there any circumstance where I have to redo this - simple and fast - procedure?

/Per

if you change the custom WB for something else this one will be lost, but it takes 10 seconds to shoot a dark frame

You do not have to keep the black frame and you are free to format the card or change your WB setting to AWB, Daylight, Tungsten, Kelvin etc. and back to Custom with no problem. Obviously, if you shoot a new custom WB image and select it then that will wipe out the black frame WB, but it certainly is quick and easy to reshoot another one.

Thanks, now I got it all! So... can you also fix some nice weather to shoot in? I want to try this out now!

/Per

Just wanted to chime in that the severley overexposed "pure white" frame works for the 7D.

I almost understand how the UniWB may work (thanks to Guillermo and others, mostly). But here are two questions:

1. Is there a quick method (akin CWB off black/saturated frame) that can be used in cameras that do not measure CWB off the recorded images? For instance, S95 can set the CWB, but only using the object it is pointed at. I triedto do it in total darkness, and it sets a greenish CWB but the multiplication coefficient are nowhere close to 1's.

There is a related idea that perhaps someone more experienced could evaluate: S95 (and many other cameras) can hand-adjust B<-->A, M<-->G colour shifts on top of prerecorded WB settings (for which the multiplication coefficients are known and recorded in EXIF). Is there a way of calculating a combination of WB preset and the colour shift to end up with a UniWB for the camera?

This would be convenient and repeatable, but I'm not sure if the range of colour shift adjustments is sufficient for such a trick. BTW how do those shifts translate to the multiplication coefficient - all Canon says is that colour can be adjusted by +/-9*7mireds - I don't really know what it means.

2. I think someone mentioned the issue before: the S95 has a life histogram, but it's the luminosity histogram (i.e the weighted average of gamma transformed R,G, and B values, which, UniWB or not, won't show clipping of 1 or 2 channels - it will show clipping if it's pure white). The Review image has 3 separate R,G, and B histograms, and these will show individual channel clipping with UniWB, but they are 'review', not 'life'. Is there a way around it?

Cheers.

tomczak wrote in post #11216907
I almost understand how the UniWB may work (thanks to Guillermo and others, mostly). But here are two questions:

1. Is there a quick method (akin CWB off black/saturated frame) that can be used in cameras that do not measure CWB off the recorded images? For instance, S95 can set the CWB, but only using the object it is pointed at. I triedto do it in total darkness, and it sets a greenish CWB but the multiplication coefficient are nowhere close to 1's.

There is a related idea that perhaps someone more experienced could evaluate: S95 (and many other cameras) can hand-adjust B<-->A, M<-->G colour shifts on top of prerecorded WB settings (for which the multiplication coefficients are known and recorded in EXIF). Is there a way of calculating a combination of WB preset and the colour shift to end up with a UniWB for the camera?

This would be convenient and repeatable, but I'm not sure if the range of colour shift adjustments is sufficient for such a trick. BTW how do those shifts translate to the multiplication coefficient - all Canon says is that colour can be adjusted by +/-9*7mireds - I don't really know what it means.

2. I think someone mentioned the issue before: the S95 has a life histogram, but it's the luminosity histogram (i.e the weighted average of gamma transformed R,G, and B values, which, UniWB or not, won't show clipping of 1 or 2 channels - it will show clipping if it's pure white). The Review image has 3 separate R,G, and B histograms, and these will show individual channel clipping with UniWB, but they are 'review', not 'life'. Is there a way around it?

Cheers.

Just use an oversaturated exposure, so that you saturate all three channels, i.e. a pure white frame. Setting at as your custom WB will force the coefficients of future exposures to 1.

Thanks. I don't know how to do 'saturated exposure' in S95 while setting the CWB, though. This is because the camera don't record the image that can be used for setting CWB. Instead, it allows for pointing to a 'gray' object and setting the CWB right then and there. The feature is automatic, i.e. I don't know how the camera controls the exposure while setting the CWB - doing it in total darkness did not result in multipliers of 1; pointing it to the sun does not either. Changing the ISO, aperture and shutter speed don't seem to affect the Custom WB feature.

tomczak wrote in post #11217505
Thanks. I don't know how to do 'saturated exposure' in S95 while setting the CWB, though. This is because the camera don't record the image that can be used for setting CWB. Instead, it allows for pointing to a 'gray' object and setting the CWB right then and there. The feature is automatic, i.e. I don't know how the camera controls the exposure while setting the CWB - doing it in total darkness did not result in multipliers of 1; pointing it to the sun does not either. Changing the ISO, aperture and shutter speed don't seem to affect the Custom WB feature.

You need to manually set your shutter to something long, and your aperture as wide as it can go, then photograph something bright and evenly lit. The sun could cause damage so I don't recommend it as a light source. Take the exposure, then set that as your custom WB.

Thank you. Setting S95 to manual, setting long exposure and than taking the CWB reading off sth. bright worked fine. I got all 3 WB multipliers very close to 1.

Now, with UniWB set, I have a question/comment: S95 (and many others) has a live histogram which shows luminescence (a weighted average of R, G and B channels, where B and to the lesser degree R contribute much less to its value than G). With the UniWB set, the histogram will show the clipping of the RAW if it is clipped in all 3 RGB channels. The RAW red and blue can be easily clipped without luminance showing it at all.

The only way I can see to use UniWB as intended (i.e. optimizing RAW exposure without clipping any channels) for the cameras that have single live histogram showing luminance only (but being able to show the individual R,G and B histograms in review), is not to use live histogram at all, but rather try to adjust the exposure based on the single channel histograms of the review of the previous shot ((which will show clipping of the individual RAW channels) and use it to adjust the next iteration of it. I wonder if using AEB wouldn't be faster and more convenient in the first place?

I'm dropping in because, coming on toward the end of '10, I thought it would be fitting to just give a tribute to panopeeper, the author of RawAnalyze, who passed this year...

Is UniWb feasible in chdk enabled cameras?

What is the general steps involved?

(eg. download freeware software, then one underexposed lens cap on shot, one overexposed flash on white paper shot, then process to find the darkest and brightest values...)

I don't think it should make any difference. Depending on the camera you may get away with either using a shot with the lens cap on (easiest) or blowing out everything (harder to get right).

There's no processing that needs to be done. Simply use the black image or white image when setting your custom white balance. If it works, then normal images will have a green tinge to them and you can check how close to unit values you are using some of the tools mentioned earlier in the thread.

Some cameras will balk at using images that are completely black or blown out. In that case you will need to use more sophisticated methods.

e-k