Should one undistort each set of calibration images with camera calibration intrinsics before stereo calibration of a pair of cameras?

edit

Question is mostly stated in the title. I have pairs of calibration images, one from each camera. I run camera calibration on the detected chessboard corners in these images, once for the left camera and once for the right camera.

Before stereocalibration, should I now undistort these images and remap the chessboard coordinates using the intrinsics learned from camera calibration? Would this improve the stereo calibration?