Dimensionality of Image transforms
The
computations for can also be reduced by restricting the choice of
to fast transforms.This implies that has a structure that allows factorization of the type,
=
where
,...... are matrices with just a few non zero entries say
where
.
Therefore a multiplication of the type :
=
is accomplished in
operations.
For several transforms like Fourier, Sine, Cosine, Hadamard etc,
,
and operations reduce to the order of
or
(for images).
Depending on the transform, an operation is defined as 1 multiplication and 1 addition or, 1 addition or subtraction as in Hadamard Transform .
Kronecker products:
If and are and matrices we define Kronecker product as:
Consider the transform,
=
or,
|
(4.1.7) |
If and denote and row vectors of and then (7) becomes ,
=
=
where is the block of
If and are row ordered into vectors and respectively, then = = ( ) .
The number of operations required for implementing equation(4.1.7) reduces from to .
|