The form of the Q matrix is given as follows:
In that image, cx and cy are the coordinates of the principal point in the left camera (if you did stereo matching with the left camera dominant), c’x is the x-coordinate of the principal point in the right camera (cx and c’x will be the same if you specified the CV_CALIB_ZERO_DISPARITY
flag for stereoRectify()
), f is the focal length and Tx is the baseline length (possibly the negative of the baseline length, it’s the translation from one optical centre to the other I think).
I would suggest having a look at the book Learning OpenCV for more information. It’s still based on the older C interface, but does a good job of explaining the underlying theory, and is where I sourced the form of the Q matrix from.