Camera and projector calibration
I have a question about camera calibration. I've followed approach shown in a book Learning OpenCV for camera calibration process. But the calibrated(undistorted) image is worse than the original one. Is it possible that my camera don't need calibration anymore?(means that the calibration is done by some driver or something like that)? In fact it seems that the original image is not distorted at all. I know that it's not only about distortion, but what would you recommend me to do?
Thanks for every reply
The calibration cannot be done by the driver. If you're planning 3D reconstruction, then you cannot do without a calibration matrix.
Several things to check:
Have you verified the corner finder is working correctly (draw them on the calibration images and see if there are any outliers)?
Taken good calibration images? Visual inspection for blur, drastic perspective changes, etc.
Have good coverage of the image area with your calibration images? The calibration is done using a nonlinear optimizer which depends heavily on inputs. I usually will take one that has target taking up entire image, four with target taking up each quadrant, then another four with target taking up ~1/9 the area in each of the four corners. Note: bias to corners is often useful since lens distortion is usually bad at the corners and minimal in the middle.
Also, have you checked the other solved intrinsics such as focal length and thought about whether it makes sense? You can compare it to what the lens says. You can also estimate it manually by estimating FOV with placing known objects at the edge of FOV and looking at their configuration relative to the camera.