1
\$\begingroup\$

I want to stitch images using equal cameras(Point Grey GS3-U3-41C6C-C) but with different lenses. One lense is a wide-angle lense (Sunex DSL318B-650-F2.4) and the other is a fish-eye lens (Fujinon FE185C086HA-1). This setup is from the 360° camera of facebook. The overlapping area for the stitching is more that enough so there should be no concern.

However, the images look quite different in terms of the color response and our stitching results looks unappealing due to the change of color.

What we did is to calibrate all cameras, i.e. wide angle and fisheye using the Macbeth ColorChecker using 3 different color temperatures. Our software-pipeline looks as follows:

  1. Blacklevel correction
  2. Whitebalance
  3. Colorcorrection
  4. Gamma correction

Hardware parameters were set to default values.

Is there a possibility to keep the color responses across the different lenses more consistent using hardware settings, e.g. hardware white balance, shutter speed, brightness etc.?

My suspicion is that we have to do adjustments for every scene that we shoot, since in the calibration process the results looked quite consistent and in real-life scenarios the color differences were clearly visible.

Here an example of what i mean:

This is the stitch result of the wide angle cameras. stitchresult-sidecameras

Fish eye projection, the resolution is lowered but that is not part of concern i guess. Projected image by fisheye

The fisheye result looks definitly warmer which leads to an inconsistent result.

\$\endgroup\$
2
  • \$\begingroup\$ Did you set both to a Custom WB measured with the lens they are using? \$\endgroup\$
    – Itai
    Commented Nov 20, 2017 at 17:39
  • \$\begingroup\$ No, so far all cameras have the same default WB value and the SW-pipeline remained the same. \$\endgroup\$
    – Daniel R.
    Commented Nov 20, 2017 at 17:54

2 Answers 2

1
\$\begingroup\$

It is possible to create a very close look out of mismatched lenses if you make color profiles for each one.

The process involves shooting a color chart with known color values in a controlled environment and determining the deviation from those known values to the actual result on the camera. You can even use the macbeth color checker to create a basic profile, but a chart with a lot more colors will allow you to craft a more accurate profile. A cheap it8 chart would be a good start, if you can afford it get a ColorChecker Digital SG

Once the profile is created you apply it to the images to have matching results.

Note that the profile it will only be valid for a specific lens and the specific color temperature used to shoot the chart.

Read: http://www.steves-digicams.com/knowledge-center/profiling-a-camera-with-an-it8-target and https://ninedegreesbelow.com/photography/well-behaved-camera-profile.html and https://photography.tutsplus.com/tutorials/how-to-create-a-custom-color-profile-for-your-camera--cms-24339

\$\endgroup\$
2
  • \$\begingroup\$ Related: photo.stackexchange.com/questions/60644/… \$\endgroup\$
    – user39557
    Commented Nov 23, 2017 at 17:10
  • \$\begingroup\$ That's what we did partly... So calibrating each camera (lets say 17) for a couple of different light temperatures (lets say 5) would result in 85 different config files? But will this calibration hold when the light temperatures differ, e.g. the light temperatur outside is 1000K colder or warmer than the color temperature we calibrated? \$\endgroup\$
    – Daniel R.
    Commented Nov 24, 2017 at 14:22
0
\$\begingroup\$

Each lens will produce chromatic abberation, and the abberation has to do with how incoming light is "bent" as it passes through the optical elements. So, the short answer is yes. Lenses with different optical elements will produce images with different chromatic abberation. If you are stitching multiple images into a single image, a best practice would be to use the same lens AND exposure settings for each pre-stitched constituent image. If you are using lenses with good chromatic abberation data, you should be able to use constituent images made using different lenses, but chromatic abberation correction is never perfect. In other words, you will never be able to acheive zero chromatic difference between constituent images made with the different lenses.

\$\endgroup\$
4
  • \$\begingroup\$ But looking at the images, this is definitely more than just chromatic aberration. There's a very clear white balance difference with one lens producing much warmer tones than the other. \$\endgroup\$ Commented Nov 23, 2017 at 19:56
  • \$\begingroup\$ I can only see a small horizontal strip of the pictures (I'm viewing this on a cell phone), but generally what I stated is applicable. \$\endgroup\$ Commented Nov 24, 2017 at 6:43
  • \$\begingroup\$ I think @user1118321 spots the problem quite well. We use whitebalancing coefficients that we calculated with 3 indoor light temperatures (3100K, 4400K and 5650K). We encountered the problem everytime when we were outside and have 'natural' light. \$\endgroup\$
    – Daniel R.
    Commented Nov 24, 2017 at 14:16
  • \$\begingroup\$ Outside with natural light in addition to artificial lighting? The color of natural light isn't actually fixed, as you may have learned in a physics course. It is dependent on various factors, such as the angle of the sun to the earth at your latitude, pollutants and other particals suspended in the atmosphere, etc. The statement of "the color balance" of natural light (from the sun) is a simplification. \$\endgroup\$ Commented Nov 24, 2017 at 16:44

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.