open3d.geometry.PointCloud.create_from_rgbd_image needs to consider lense distortion

I am using Kinect Azure for the depth sensor, and found the PCD output from Kinect SDK and the results from open3D is different.

The main reason is that Kinect SDK uses lense distortion correction and open3d only uses fx, fy, cx, cy for Camera intrinsic. The results are dramatic when we work with small object 3D reconstruction case.

Please extend the Camera Intrinsic model and add the option for lense distortion correction.

Please do it for not being just a toy project.