Kronecker product and SVD approximations in image restoration (Q1124780)

A linear discrete model of image restoration is defined by the matrix-vector equation \(g= Kf+ n\), where \(g\) is the known observed image, \(f\) is the actual unknown image, \(n\) is an additive noise and \(K\) is a banded block Toeplitz matrix with banded Toeplitz blocks. In order to obtain an estimate of \(f\), the authors replace \(K\) by its singular value decomposition. Comparison with Tikhonov approximation and illustrative examples are provided.