Manually setting the colour temperature

Submitted by Richard Yorke on Thu, 11/13/2014 - 12:43
Forums: 
General discussion

The new beta is showing good progress and I have been trying it out on a wider assortment of pictures. While doing this I have been including pictures taken with 2 Canon cameras using the CHDK hack, an Ixus 70 and a PowerShot SX10 IS. Neither of these are supported and I wonder if this is why the manual setting of the colour temperature is unavailable, and I would be interested in :-
1) What other limitations apply to unsupported cameras.
2) Is it easy to add cameras to the supported list and how do we go about it.

lexa

Thu, 11/13/2014 - 14:20

There is no way to calculate CCT/Tint WB without color profile data. Also, there is no way to show raw image right without color profile.

So, to support new camera(s) we need



 a) Sample file(s) to get basic file reading support in (our) LibRaw library. CHDK-files should be easy to add, but I need at least one sample with known colors (e.g. sky) and with deep shadows (to approximate black level data) for each camera.  Sample files are necessary but not always sufficient, because for some formats compession scheme is still unknown for us.



 b) Color target shot(s) (Colorchecker SG is preferred), made under even light with good spectrum (ImagingResources files are good example: http://www.imaging-resource.com/PRODS/canon-7d-mark-ii/E7D2hVFAI00400.CR... ). Standard 24-patch color checker is not enough to build good color profile. The target shot should be right exposed, slightly defocused and without hotspots/glare.

Also, it is better to have two shots for color profiling 1) flat gray surface  and 2) color target put on this surface. It will help us to correct for light uneveness.

 

--

Alex Tutubalin/FastRawViewer team

Iliah

Thu, 11/13/2014 - 15:28

To add a bit, a good start is to have pairs of chdk/jpg images of just about anything, taken at all major white balance presets in camera: Incandescent, Flash, Fluorescent, Sunny, Shadow, Cloudy. That _may_ help us populating the drop-down for WB presets, which IMHO is a step forward; however the chdk data may not have sufficient information to guess the camera model. In that case we may be stuck. But let's try and see first.

Richard Yorke

Thu, 11/13/2014 - 17:03

OK, that is very helpful. I should say that even without the profiles the images don't look bad and the contrast curves and exposure correction work, as do the focus peaks, so that a reasonable assessment of the image is perfectly possible without a camera profile, though the more profiles you have available the better.
As for creating profiles, one the CHDK data is amalgamated with the jpg header into a DNG file with 'DNG4PS' the camera model is in the EXIF data so that will not be a problem. I was about to ask about what images you would like when your additional bit arrived. Taking those shots should not be a problem, other than finding the time. However, colour target shots may be a little harder as those downloaded are only as good as the printer, or screen, and as my screen is set up better than my printer and has a bigger gamut, I was going to suggest screen shots might be the best option without the (quite considerable) expense of buying an accurate printed chart. If all that sounds reasonable I will put together a collection of images for you, though there is no great urgency for this.

Iliah

Thu, 11/13/2014 - 17:23

I think a study of WB presets may help figuring out the sensor, and the colour transforms ("profiles") may be in that case borrowed from a camera using the same sensor, but with the known CFA properties.

lexa

Fri, 11/14/2014 - 00:58

Unfortunately, the only real way to do 'Colorchecker SG shots' it to shot real ColorChecker SG (bought from X-Rite).

Screen emulation photo (or print emulation photo) will not work, unless you measure real spectral response of your screen/print using spectrofotometer.

--

Alex Tutubalin/FastRawViewer team

Richard Yorke

Sat, 11/15/2014 - 12:27

Yes I fully realise that really the proper chart is really needed, I just wondered what might be a rough alternative as I don’t want to spend the sort of money they cost just to enable manual adjustment via the sliders (though I can use Alt-left mouse click without a profile).
 
Anyway, I have done a little research and found that the Canon Ixus 70 would seem to use the same sensor as the Canon Powershot A470 and A550 which you have a profiles for, Changing the EXIF of a picture taken by the Ixus to indicate it was taken by a PowerShot A470 results in a reasonable result with Daylight, Auto, Custom and UniWB being available and As Shot and Tungsten greyed out. But most importantly the sliders are available. Perhaps you could direct images with Ixus 70 in the EXIF to the A470 profile. I did go so far as to take some shots of a screen chart with the Ixus in case you needed anything else from the files but adding the link triggered a spam warning :-(
 
The Powershot SX10 does not share a sensor with any of the cameras you have a profile for, though making FRV think the images are from a Powershot SX1 which has a sensor of the same size and resolution but a different type, does also allow the sliders and presets to be used. If this is enough for you to direct the SX10 to the SX1 profile that seems fine, though if you want similar images taken by that camera I can always do that for you. I would then just have to work out how to get the links to you.

lexa

Sat, 11/15/2014 - 12:31

Could you please provide several samples for these cameras?

(Dropbox upload and send link to me or just share for lexa@lexa.ru is OK)



 

--

Alex Tutubalin/FastRawViewer team

Add new comment