CODEWITHSUNDEEP
graphicscomputer-visioncamera-calibration
Carl
Carl

Reputation: 105

Camera calibration algorithm evaluation

Recently, I am developing algorithm to improvement the camera calibration algorithm in my research group. I would like to ask are there any method to evaluation the camera calibration algorithm so that I can compare the result among difference algorithm?

The most easiest way I can think of is taking the mean square average of different between the calibrated one and the original one pixel-wise. Are there any other suggestions?

Upvotes: 1

Views: 1091

Answers (1)

Dima
Dima

Reputation: 39419

Are you talking about geometric camera calibration (focal length, optical center, etc.), or color calibration?

For geometric camera calibration, the main criterion is reprojection errors. Presumably, you are using some sort of calibration pattern, like a checkerboard, where you can detect a set of points. To evaluate calibration accuracy, you look at the distances between the detected points and the reprojected points. This is what a calibration algorithm typically tries to minimize. Depending on the calibration software you use, you may also be able to look at the uncertainty of the estimated camera parameters. See this example in MATLAB.

Alternatively, you can use your calibrated camera to measure an object of a known size, and see how precise your measurement is.

Upvotes: 1

Related Questions