When calibrating a camera with a chessboard, are more squares better or worse? When would you use more or less? And how does the size of the blocks effect the calibration results?


Generally speaking, when you are estimating a model from data, the more (independent) data points, the better the estimation.

In this particular case, the goal is to get as many 3d -> 2d point correspondences as possible, filling as much as possible of the volume of space of interest that is observed by the camera. The size of the squares directly affects how many data points you get per image (larger squares implies less visible corners/points).

It is also important to ensure that the camera parameters you want to calibrate are observable given the data. For example, in order to calibrate the focal length (or the field of view, equivalently) you need to observe a significant perspective foreshortening of the (planar) checkerboard in a few images. In order to accurately calibrate the nonlinear (e.g. barrel) lens distortion, you'll need to observe plenty of aligned data points near the edge of the image, where the distortion is larger.

There is a longer response on StackOverflow that you may want to look up.

  • $\begingroup$ I have done the experiments and I can confirm this. More squares = more tracking features and therfore more accuracy. But obviously higher resolution is needed when detecting the features from further distances, because more the features become "too close together" from the perspective of the camera. $\endgroup$ – Grim Dec 7 '14 at 23:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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