4 Replies Latest reply on Nov 12, 2009 10:28 AM by Bill_Janes

    How to Preserve Colors

    rcholman7

      I have a set of photos taken of different color objects illuminated by different types of lights at night along with a photo of each object in daylight.  Color temp of the lights ranged from 2000K to 6000k. I want to show the difference in color rendering between the different light sources.  I don't want to "improve" the photos -- just show the color distortion as it was -- that's the point.  Photos were taken on a Nikon D50 SLR in raw with auto white balance.

       

      My question is:  I've used "as shot" white balance when importing into Photoshop.  Is that correct?  I'm concerned that the white balance computed by the camera will be used and will distort the original colors.

        • 1. Re: How to Preserve Colors
          Bill_Janes Level 2

          rcholman7 wrote:

           

          I have a set of photos taken of different color objects illuminated by different types of lights at night along with a photo of each object in daylight.  Color temp of the lights ranged from 2000K to 6000k. I want to show the difference in color rendering between the different light sources.  I don't want to "improve" the photos -- just show the color distortion as it was -- that's the point.  Photos were taken on a Nikon D50 SLR in raw with auto white balance.

           

          My question is:  I've used "as shot" white balance when importing into Photoshop.  Is that correct?  I'm concerned that the white balance computed by the camera will be used and will distort the original colors.

          The "as shot" will produce neutral whites only if the auto white balance on the camera is working properly. My experience with Nikon Digital cameras (the D70, D200 and the D3) is that WB works well for the higher color temperatures but falls short below 3000K or so, white in the scene will be rendered with an orange cast. In the latter situations it would be best to use a preset white balance by taking a reading from a white card.

           

          Your comment about showing the "color distortion" suggests that you may not want to render white as white in the photograph. In other words, you may want tungsten light to be yellow or sunset light to be reddish. Then auto white balance will not do the job in that case either, since it is attempting imperfectly to render white as white in the photograph. This brings up the subject of color adaption in human vision. The visual system can adapt to 3200K light so that white appears neutral, but at very low color temperatures adaption is incomplete. However, when you view a photograph the eye is adapted to the ambient light, not white in the photograph.

          1 person found this helpful
          • 2. Re: How to Preserve Colors
            rcholman7 Level 1

            Thank you Bill.  Good point about adaptability of the human eye.  I think you've also confirmed that "as shot" is not what I want.

             

            I'm still looking for a solution.  As you point out, I'm NOT looking for whites to be rendered as neutral white in the finished photos.  Instead, I'm looking to render whites (and several other colors) at the same distorted color captured by the camera due to differences in the spectral composition of the light used for the scene.  I'm actually trying to show the difference in color rendering for each type of light as compared to a daylight photo.

             

            I measured the color temperature of the light used to illuminate the object (at night) with a separate light meter, so I know very precisely what the color temperature is of the light used for each photo.

            • 3. Re: How to Preserve Colors
              Panoholic Level 2
              I'm looking to render whites (and several other colors) at the same distorted color captured by the camera due to differences in the spectral composition of the light used for the scene

               

              There is no such thing as "preserving the colors" but "reproducing the colors", because the camera's way of recording the image is different from what you get at the end (already the thread title How to Preserve Colors showed this misconception).

               

              The pixels of a camera with Bayer type sensor are often described as "red", "green" and "blue". This is comfortable, but incorrect; the pixels should be called Rgb, Grb, Brg, i.e. "mainly red, somewhat green, somewhat blue", the other one "mainly green, somewhat red, somewhat blue" and the third one is "mainly blue, somewhat red, somewhat green".

               

              Let's see an example created with your camera (the sample is from Imaging Resources). We concentrate on the red, green and blue patches of the color checker card. The captures show a crop from the raw, i.e. undemosaiced, non-WBd image. The first capture shows the image in "composite color" mode: the "red" pixel value was used for the R component of RGB, the "blue" pixel value for the B component, and the average of the two "green" pixels became the G component.

               


              http://www.panopeeper.com/Nikon/NikonD60_RGBDemo_Composite.GIF

              The patches appear blue, green and red, more or less. But how are they really? The following captures show only the red, the green and the blue components. If a patch were red, then the green and blue component would be null (black in the display), likewise with the green and blue. Is that so?

               

              http://www.panopeeper.com/Nikon/NikonD60_RGBDemo_Red.GIF http://www.panopeeper.com/Nikon/NikonD60_RGBDemo_Green.GIF

              http://www.panopeeper.com/Nikon/NikonD60_RGBDemo_Blue.GIF

              It is obvious, that the "natural" color component is the strongest, but the others are far from being black.

               

              The next issue is, that this appearance does not depend only on the color of the object and on the illumination, but on the sensor (on the filters over the sensels) as well. Look at the same patch under the same illumination, created by the D90. It is close, but not identical to the D60:

              http://www.panopeeper.com/Nikon/NikonD90_RGBDemo_Composite.GIF

              Again, all the above was without white balancing.

               

              If you want to show, how the D60 (but only the D60!) records a scenery under a givel illumination, you have to use a neutral white balance. First, create such a template shot. The easiest way (though not all cameras like this) is to make a shot of something, which totally clips, like the cloudy sky or a white sheet, at least three stops overexposed (do NOT shoot the sun). Then use this image for setting the custom WB in the camera and select "custom WB" for the demo shot. Now the colors are not adjusted to the illumination.

               

              This is still far from what the camera captured, but somewhat closer. If you want to see really the captured data, use Rawnalyze (free): http://www.cryptobola.com/PhotoBola/Rawnalyze.htm

               

              Gabor

              • 4. Re: How to Preserve Colors
                Bill_Janes Level 2

                rcholman7 wrote:

                 

                Thank you Bill.  Good point about adaptability of the human eye.  I think you've also confirmed that "as shot" is not what I want.

                 

                I'm still looking for a solution.  As you point out, I'm NOT looking for whites to be rendered as neutral white in the finished photos.  Instead, I'm looking to render whites (and several other colors) at the same distorted color captured by the camera due to differences in the spectral composition of the light used for the scene.  I'm actually trying to show the difference in color rendering for each type of light as compared to a daylight photo.

                 

                I measured the color temperature of the light used to illuminate the object (at night) with a separate light meter, so I know very precisely what the color temperature is of the light used for each photo.

                As I now understand the problem, you want the spectral distribution of light reflected by objects in scene illuminated by light with a given color temperature to match those in the image without any chromatic adaption. To simplify, the amount of red, blue, and green reflected by a white object in the scene would be represented by the same RGB values in the image. Of course, red, blue and green have to be quantified, say by the primaries in sRGB.

                 

                The spectral response of the D50 sensor can be quantified, for example in this DXO link (click on the color response tab):

                 

                http://www.dxomark.com/index.php/eng/Image-Quality-Database/Nikon/D50#

                 

                You could use the color matrix to try to reproduce the scene colors without any chromatic adaption, but this is beyond my expertise. Perhaps Eric Chan or other experts could comment. Of course, all the camera could do is produce a tristimulus match--the actual spectral colors in the scene could not be reproduced. You may not know the color temperature of the ambient light as accurately as you think, since color meters work best with light produced by a black body radiator.