OpenCV 3 Computer Vision with Python Cookbook

How it works

cv2.decomposeHomographyMat decomposes the homography matrix into rotations and translations. Because the solution isn't unique, the function returns up to four possible sets of translation, rotation, and normal vectors. cv2.decomposeHomographyMat accepts the 3x3 homography matrix and 3x3 camera matrix as arguments. The return values are: the number of found solutions, a list of 3x3 rotation matrices, a list of translation vectors, and a list of normal vectors. Each returned list contains as many elements as the number of solutions that have been found.