How can I identify the projection type of a fisheye lens and estimate its distortion?

Asked 7/21/2017

6 views

2 answers

I have a Fujinon FE185C057HA fisheye lens and want to determine what projection it uses (for example equidistant, stereographic, equisolid, Brown, or Baker). I have the lens available, so I can shoot test images if needed.

What is a practical way to test this from photos? Are there software tools that can help identify the fisheye projection, and if the lens follows a calibrated distortion model such as Brown or Baker, estimate the distortion parameters?

opencv distortion fisheye hugin lens-calibration

Originally by Photography Stack Exchange contributor. Source · Licensed CC BY-SA 4.0

Photography Stack Exchange contributor

9y ago

2 Answers

I realize the question is quite old, but it still seems to get some views.

The quickest thing that I would try first is to take a photo of a planar grid / checkerboard pattern / brick wall straight on (center of image should be perpendicular to grid surface). Then I would use Hugin to see if I can undistort (de-fish) the image to obtain the original grid pattern with reasonably straight lines.

Looking at the lens type, you may use it for an engineering application. One good option to calibrate such a lens is using OpenCV. I found a tutorial how to calibrate fisheye lenses with OpenCV.

Originally by user107563. Source · Licensed CC BY-SA 4.0

user107563

3y ago

AI Answer

Generated from our catalog & community — verify before relying on it.

A practical first test is to photograph a flat, regular pattern straight on—such as a checkerboard, grid, or brick wall—with the lens axis centered and perpendicular to the surface. Then try calibrating or “de-fishing” the image.

Two useful tools mentioned are:

Hugin: good for testing whether the image can be undistorted into a grid with straight lines, which can help you infer the projection model.
OpenCV: a stronger choice if you need an engineering-style calibration. Its fisheye calibration workflow can estimate lens parameters from multiple images of a checkerboard.

If you specifically need distortion coefficients, OpenCV is the better route because it is designed to solve for calibration parameters rather than just visually correct the image.

So the quick workflow is: shoot a checkerboard/grid carefully, test in Hugin for a fast projection check, and use OpenCV calibration if you need actual distortion parameters.

How can I identify the projection type of a fisheye lens and estimate its distortion?

2 Answers

Your Answer

Related Questions