I have checked and examined OpenGL documentation to construct intrinsic and extrinsic matrices from projection and view matrix. However, I could not find a clear conversion. What I have found up to now is two of three components in intrinsic matrix in pybullet by making extensive amount of calculation and research:

Intrinsic Components:

------------------------------

* Focal length values (fx, fy)

* Two principal point values (cx, cy)

* Pixel Skew (s)

What I found:

------------------

* Focal Length: Since there is only one camera eye, there is only one focal length, which is equal to 1 / tan(q/2) in which q refers to the angle "field of view" in computeProjectionMatrixFOV function.

* Principal Values: The principal points are indices of middle pixel in the image, which is 250 for 500x500 image.

**However, In spite of my 2 days research, I could not find pixel skew value. Is there anyone who can provide pixel skew and extrinsic matrix of pybullet camera image ? I think that we need these matrices for many computer vision purposes.**