projection
A linear transformation of a vector space is called aprojection if it acts like the identity
on its image. Thiscondition can be more succinctly expressed by the equation
(1) |
Proposition 1
If is a projection, thenits image and the kernel are complementary subspaces, namely
(2) |
Proof. Suppose that is a projection. Let be given, and set
The projection condition (1) then impliesthat , and we can write as the sum of an image andkernel vectors:
This decomposition is unique, because theintersection of the image and the kernel is the trivial subspace
.Indeed, suppose that is in both the image and the kernel of .Then, and , and hence . QED
Conversely, every direct sum decomposition
corresponds to a projection defined by
Specializing somewhat, suppose that the ground field is or and that is equipped with a positive-definite innerproduct. In this setting we call an endomorphism an orthogonal projection if it is self-dual
in addition to satisfying the projection condition (1).
Proposition 2
The kernel and image of an orthogonal projection are orthogonal subspaces.
Proof. Let and be given. Since is self-dual we have
QED
Thus we see that a orthogonal projection projects a onto in an orthogonal fashion, i.e.
for all .