How to decompose the "camera projection matrix" into camera internal and external parameters?

That is, what conditions are needed to proceed? Can the internal and external parameters of the camera be reversed by using n (n>6) non-coplanar corresponding world coordinate points of the image?

related issues: https://www.mathworks.com/matlabcentral/answers/472171-how-to-calculate-the-camera-intrinsics-k-rotation-matrix-r-and-translation-vector-t-through-the-ca

No one answered so far.

Answer 1

[Matt J]

If you have the 3x4 camera projection matrix already, then it can be directly decomposed into intrinsics and extrinsics. The attached file will do it, e.g.,

P=rand(3,4);  %A projection matrix

[K,R,ImC]=Pdecomp(P)

K = 3×3

0.4803 -0.1453 0.7873 0 0.6865 0.9392 0 0 1.4118

R = 3×3

0.2997 -0.7858 0.5410 -0.6539 0.2437 0.7162 0.6947 0.5685 0.4408

ImC = 3×4

1.0000 0 -0.0000 0.7352 -0.0000 1.0000 -0.0000 -0.5784 0.0000 -0.0000 1.0000 0.5442