undestored image or original one as input?

[GandalfTGrey]

After the intrinsic calibration process, getting the intrinsic $K$ and destortion parameters $\sigma$, Should i use the undestorted images $\hat{I}$ and camera intrinsics $K$ as the ILCC input， or use the original image $I$ and $K$?