0
\$\begingroup\$

enter image description hereI took a photo of this stuff with my phone. It is for a Wikipedia article.

The stuff is off-white, or more precisely white with orine-ish tint. Instead, I got a photo using my phone where it is light gray and the white background is rendered light cyan in color.

Why is that, and do you think I can manipulate the image so that the powder is off-white with oringish tint and the background is pure white, and if not, what have I done wrong?

Oh, it is a powder, btw.

And notice how the color of the background paper is cyan-ish instead of white, and I think that is the negative of orange. This is why I believe this image can be shifted to display the accurate colors. But I don't know of course, that is why I ask your professional advice

\$\endgroup\$
15
  • \$\begingroup\$ Could you not share the image? \$\endgroup\$
    – MrUpsidown
    Nov 24, 2021 at 16:35
  • \$\begingroup\$ Don't know if that is sarcasm, but also don't know where it will be appropriate to upload the image \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 16:39
  • 1
    \$\begingroup\$ Why would it be sarcasm? It's a whole lot easier to give advice if we know what we're advising on. You can upload to your question if you make sure the file is 2MB or less. It won't let you inline it visibly, but a higher rep user can do that for you afterwards. \$\endgroup\$
    – Tetsujin
    Nov 24, 2021 at 16:51
  • \$\begingroup\$ ok, my bad. Could you recommend a good and free and opensource JPG metadata remover tool for Windows? I will upload as soon as I find one \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 16:56
  • 1
    \$\begingroup\$ figured it out,sorry. so basically that is it. Need it to be the correct color for wikipedia. And I was thinking that if the powder is not the correct color, and so is the white sheet of paper behind it, I might be able to shift the colors until the paper gets snow-white and the powder will turn the actual color. Probably with GIMP because I have it already \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 17:12

3 Answers 3

1
\$\begingroup\$

You have two problems when taking photos like this and allowing the camera to automatically choose the settings for white balance and exposure. They're both related to the same basic assumption cameras make when set to "Auto" everything:

Unless told otherwise, a camera will assume that the predominant thing in the scene is a neutral color of medium brightness. In other words, it assumes that whatever dominates the frame is medium gray, halfway between black and white.

When you put something that is not a neutral shade of white in most of the camera's field of view, the camera's automated routine will assume the lighting is creating a tint and will attempt to adjust WB to compensate. Various different light sources have different color temperatures and tints. Cameras have to adjust for this to render colors that look correct to our eyes. Our eye/brain vision system also automatically adjusts for a vey wide range of light sources.

Since your powder, as well as the white background, is fairly light in color the camera underexposed the shot to make it fall halfway between the brightest thing it can record (pure white) and the darkest thing it can record (pure black). To compensate for this you need to tell the camera to expose brighter than it otherwise would. We call this exposure compensation, though a more accurate name might be "light meter bias" or "meter calibration", because that's what exposure compensation actually does.

Since your powder has a slight yellowish/orange tint to it, the camera compensated in the opposite direction on the color wheel and increased amplification in the blue channel by a lot while also slightly reducing the red channel.

The best way to deal with this is to shoot under a light source with a known color temperature and tint and set the camera's WB manually. If you're under Tungsten lights, use the camera's "Tungsten" or "Indoor" preset. If you're outside in bright sunshine, use "Daylight". Alternately, you can specify the color temperature along the blue ←→ amber axis and the tint along the green ←→ magenta axis that are more or less orthogonal to each other in color space. If you're under traditional fluorescent lighting, set the CT to around 3700K and move the tint several units towards magenta to compensate for the green tint of the lights. If you're outside in bright daylight, set the CT to 5200K and leave tint centered between green and magenta.

Your example image is tough to fix for a couple of reasons.

  • The primary one being that when a camera selects the white balance and creates an image, it only uses the information gathered by the camera that is needed to render the image this way. A lot of the other information gathered by the sensor is discarded to make the resulting image file as small as possible. You need this other information to correctly alter white balance.
  • It's also tougher to deal with because it was exposed so darkly that there is a lot of chrominance noise in it. Chrominance noise is what makes a uniformly colored background look like it's made up of lots of different specs of various colors. This is due to the limited amount of light collected and the random nature of the distribution of photons within a light field of a specific intensity. The less light one collects, the more the randomness will affect the resulting image. The more light one collects, the better the differences will average out and give a more uniform color for each of the sensor's millions of photosites.

Here's the best I could do with a quick and dirty "white balance clicker" tool. These tools are much more effective when used before converting the raw sensor data to a compressed raster image where so much of the collected information is discarded.

enter image description here

\$\endgroup\$
3
  • \$\begingroup\$ Thank you for not judging my blatant ignorance of photography, and instead of down-voting providing an actual very detailed useful answer. \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 18:31
  • \$\begingroup\$ Now that I know it is supposed to be a pink tint rather than orange, I went back and reworked my best attempt at correction. \$\endgroup\$
    – Michael C
    Nov 24, 2021 at 18:45
  • \$\begingroup\$ Well, it is supposed to be pink. I thought it was slightly orange, but it was probably very slightly pink, and I have misinterpreted it - makes more sense. Here the colors are more exaggerated and it is obvious \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 18:50
0
\$\begingroup\$

What you need is correction of white balance. In your case probably you take JPG photo and it is not so easy and in many cases not so successful (to change white balance). You can try with GIMP (this menu). Here also is another tutorial for GIMP. Also you ca try (if you phone support it) to take photo in (kind of) manual mode where you adjust it (in Android is named PRO mode). You can use Lightroom mobile (AFAIK work also on iOS) and when you take photo you have option to correct it (not sure if this option exist in free version). But with Lightroom mobile you need to export the image after take it.

\$\endgroup\$
10
  • \$\begingroup\$ Thank for the advice, but this does not seem to work in my case \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 17:28
  • \$\begingroup\$ I have posted the picture, if someone can restore it to kind of its original colors I will post it in wikipedia. If that is even possible with a JPG. I think I might need to hire a professional photographer for this one... \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 17:30
  • 1
    \$\begingroup\$ @Romeo_Ninov That is not perfect but is close enough. It will do. Thank you very much \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 17:49
  • 1
    \$\begingroup\$ I'm done with that. Here is what you have contributed to: en.wikipedia.org/wiki/Diphenylcarbazide Does not need to be perfect - this reagent gets pinker as it reacts with the oxygen in the air, so a bit more pink does not hurt. The idea is to accurately represent the chemical, and this does it. \$\endgroup\$
    – kjsdfkns
    Nov 24, 2021 at 18:12
  • 1
    \$\begingroup\$ @kjsdfkns, if someone want to read "Война и мир" :D \$\endgroup\$ Nov 24, 2021 at 18:35
0
\$\begingroup\$

With Gimp:

  1. Add one or more (up to 4) "Sample points" (Ctr-Shift-drag from a ruler) where you have a neutral color (in you case, the paper that we will assume to be white)
  2. Make the "Sample points" dialog show (Windows > Dockable dialog > Sample points)
  3. Colors > Color temperature and change the "Intended temperature" while matching the RGB values in the sample points. Move until R≈B on all the points. You will find that the Green channel isn't equal to the others
  4. With the Sample Points dialog still up, use the Levels or Curves tools to reduce the Green channel so that it aligns with the other two channels in your sample points.

enter image description here

\$\endgroup\$
3
  • \$\begingroup\$ When the neutral area is this saturated with chrominance noise it's always a crap shoot. Using several different sets of randomly selected points can still get widely varying results, though it will likely be more consistent than selecting a single point. \$\endgroup\$
    – Michael C
    Nov 25, 2021 at 9:13
  • \$\begingroup\$ Yes.What you can also do is average an area and use that color as the reference to become neutral. \$\endgroup\$
    – xenoid
    Nov 25, 2021 at 9:28
  • \$\begingroup\$ If the paper is coated with certain coatings it will also fluoresce slightly blue if the light source has a UV component (most do). Attempting to compensate for this narrow band of fluorescence can throw all of the other colors off. \$\endgroup\$
    – Michael C
    Nov 25, 2021 at 9:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.