I followed the example in this tutorial: http://www.aishack.in/2010/07/calibrating-undistorting-with-opencv-in-c-oh-yeah/ ...and I got these results...
...where the middle image shows successful 3x7 corner detection but the corrected image on the right doesn't 'flatten' the image as expected.
One deviation from the tutorial: I only sampled the checker board once instead of multiple time. The image sampled is the image on the left. A post I found online suggested that too-few samples would result in calibration problems. I can, of course, post my values in the matrices updated by the calibrateCamera() call - I wanted to start with a short post.
Will a single sample work with calibrateCamera()?
A follow-on question: Could I do a similar remap using remap() with just a 3x7 matrix of points? The examples I found that use remap() appear to specify a matrix for ALL pixels to be remapped. I'd like to try a remap by specifying where I want the 3x7 matrix of points to appear. This would allow me to skip the calibrateCamera() call and use a more arbitrary remapping.