Procedure for conducting a two-dimensional spectral transformation or spectral transformation.

摘要

It is known to separate two-dimensional spectral transforms, which transform data from the original domain into the spectral domain, into two one-dimensional transforms. Computation of the coefficients of this spectral domain is bound up with a large number of arithmetic operations. This expense can be reduced by decomposing the transform tensor [T] into a product of derived tensors [Ti], i = 1,...,a. The initial data present in an initial matrix [X] are input line-wise into a memory S. To carry out the transform, a matrix-vector multiplication is carried out between the derived tensor [Ta] and the initial vector, and the result is written into the memory S. Using the content of the memory S, a matrix-vector multiplication is carried out only (a-1) times with the tensors [Ti], i = 1,...,a-1, results of the sth matrix-vector multiplication being written once again into the memory S, and is thus a factor of the (s + 1)th multiplication. The result vector contains the coefficients of the spectral domain. It is advantageous to apply the method for image transmission systems. IMAGE