Learning dictionaries as a sum of Kronecker products. A rearrangement operation1 transforms a (sum of R) Kronecker product(s) into a rank-1 (rank-R) matrix. Inducing a Sum of Kronecker products (SuKro) structure thus becomes a low-rank approximation problem.

This code corresponds to the following paper:

C. F. Dantas, M. N. da Costa, R. R. Lopes, “Learning Dictionaries as a Sum of Kronecker Products.” In IEEE Signal Processing Letters, 2017.

It contains an image denoising demo.

Related slides here and here (from slide 14).

  1. C.F. Van Loan, n. Pitsianis “Approximation with Kronecker Products”, 1993.