4

I am doing some lab work and need to take pictures of my results. I usually take photos with any available device, even a mobilephone camera would do it for me. Lately, I had some results of solids that show a very small difference in colour (light grey and white) which you can see by eye but, unsurprisingly, cannot be detected with the phone camera. Now I am planning to buy a USB camera and connect to my computer and take some proper photos.

My question here is what property of a camera would define the sensitivity of the camera to the colours? I have found these ones online, but I am not sure if they can detect this small difference in colours.

http://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=4024

mattdm
  • 143,140
  • 52
  • 417
  • 741
Error404
  • 155
  • 1
  • 1
  • 4
  • 2
    I think probably that the problem here is actually that the phone camera is automatically adjusting the white balance and exposure so that the scene appears neutral. You need a camera with full manual control to use for what you want it for. – mattdm Jun 28 '13 at 16:30
  • 1
    Do you actually care about color or do you want to show shade of white/gray? (Or both?) – mattdm Jun 28 '13 at 16:33
  • 1
    My answer was on the assumption that you had both colors/shades in the same picture, and needed a way to tell them apart. If you are comparing across two different pictures, I'd start with mattdm's suggestion. – j-g-faustus Jun 28 '13 at 16:38
  • they are in the same picture. So iw ill depend on your answer @j-g-faustus. Cheers – Error404 Jun 28 '13 at 17:18

2 Answers2

4

There are two scenarios with different solutions:

  • You have different pictures and want to keep the color and brightness comparable from picture to picture. (Automatic exposure and automatic white balance can negate the differences you're trying to capture.)
  • You have subtle variations in colors/shades in the same picture and want to be able to tell them apart.

Consistency over multiple pictures

You need the same lighting, exposure and color balance.

Two common solutions are:

  • Use a camera where exposure and white balance (WB) can be set manually, and keep everything the same between shots, including light sources and distance between light source and subject.
  • Include a gray card in the picture as an exposure and WB reference, and use the WB adjustment frequently found in image editors. RAW is best for this (rather than JPG), as RAW has more leeway for adjustments afterwards. You'll need a camera with RAW support, which excludes camera phones and consumer webcams, but includes DSLRs, mirrorless and some compacts.
    (For scientific cameras I've no idea, but since they are explicitly scientific I can only assume that they give access to the raw data in one form or another, although you may need custom software to manipulate those data.)

Note that fluorescent lighting and sodium-vapor lamps are tricky, and can make accurate colors impossible. Avoid them if you can.

Incandescent light, photographic flashes/strobes and normal daylight are good light sources for accurate color reproduction - you want a light source that emits full-spectrum light.

Subtle variations in the same picture

I think this is primarily a matter of "bit depth".

More bits means that the sensor (or technically the ADC, the Analog-Digital Converter) can capture more nuances. One extra bit translates to twice the number of brightness levels. For an RGB sensor it's one extra bit per color and 2^3=8 times the number of distinct colors.

For reference, DSLRs and large-sensor mirrorless are typically 12-14 bits. Many/most of them can be connected to a computer, look for tethering options.
The USB cameras in your link are up to 10 bits, which may or may not be sufficient.

But: JPG and most standard monitors are limited to 8 bits, and some low-end monitors are only 6 bits. So to make the larger bit depth visible on a standard monitor, you'll need to shoot in RAW and find a mapping from a many-bits color space to a few-bits color space that retains the distinctions you want to keep.

By converting to a smaller-bit-depth color space you are inevitably discarding information. But when you define the mapping yourself, rather than leaving it to the default JPG conversion, you can decide which color and brightness distinctions to keep.

I don't know what kind of support the USB cameras offer for such remapping, but I think that's what you are looking for.
And you might be interested in a high-bit-depth monitor to go with the camera, certain high-end monitors can display 10 to 12 bits of color.

Community
  • 1
j-g-faustus
  • 3,743
  • 1
  • 22
  • 26
  • 1
    That's a perfect explanation, thanks buddy. I am showing the data in presentations later on, so I will try to get a high-bit camera to take the photos and see if they are clear enough. Many thanks. :) – Error404 Jun 28 '13 at 17:18
  • 1
    Nice answer covering both situations! – mattdm Jun 28 '13 at 19:51
  • Also for just measuring the colour of a surface rather than looking for variations like striations then do you think a colorimeter / spectrophotometer like the ColorMunki be worth looking at?
    • I've no affiliation with them, just know a few people who use them for a similar purpose to the one the OP asked about.
     – James Snell Jun 30 '13 at 23:18
  • @JamesSnell I guess it should work if you use consistent lighting. Haven't tried, so I can't really comment beyond that. But you might want one anyway to calibrate the monitor. – j-g-faustus Jul 01 '13 at 00:14
1

I think you may want to consider a Foveon based camera. The bayer sensor in most cameras has to extrapolate all colors from separate red, green, and blue filtered sensors - and it seems like because all color is interpolated, it's easy for the software to smooth out small differences in color variation. For each pixel in the output image, three spatially separate photosites (or more) were polled along with the single filtered photosite at that pixel location and combined to figure out what color is there.

The Foveon sensor, instead of using a color filter allows for different wavelengths of light to penetrate the sensor at different depths, roughly corresponding to R/G/B. This means that for each pixel in the output image there are three sensors at that exact pixel location recording different wavelengths, which is a much better system for capturing pixels next to each other that have a small variance.

The only camera maker (these days) that uses the Foveon sensor is Sigma. Current Sigma cameras all use the Merrill version of the Foveon sensor, which includes a DSLR and a set of three compact cameras of different focal lengths.

In order to test this theory, you could get a somewhat older Sigma camera that did not have the resolution of the Merrill cameras. Even the older cameras should have enough resolution to see if you can discern differences clearly.

I would look for the Sigma DP2, an older version of the current compact DP-2M camera. It should be able to be found cheaply now and you could test the effect of shooting your subject with it.

Kendall Helmstetter Gelner
  • 3,442
  • 19
  • 20