Cofactor transposition constructs the transpose of a matrix’s cofactor matrix. It represents the inverse up to scalar multiplication, avoiding explicit determinant division. section 4.3.2
For projective transformations, this is sufficient: homogeneous coordinates treat nonzero scalar multiples as the same point. Divide by the final coordinate only when normalized tuples ending in 1 are required. section 4.3.2
Undoing a composed rectangle transformation
A composed 3×3 transformation moves a rectangle’s corners. SymPy’s ordinary inverse restores the original coordinates; the cofactor-transposed matrix returns their negatives, which are equivalent homogeneous coordinates after rescaling.
