In this paper, we propose a modified orthogonal rank-one matrix pursuit algorithm based on the mean value revision. This algorithm guarantees the error matrices to be Toeplitz matrices, which decrease the SVD time and improve the precision. The error matrices are proved to converge to zero in the known element set. Finally, we show that the new algorithm is much effective in CPU time and precision than the algorithm of general OR1MP and EOR1MP through numerical experiments for Toeplitz matrix completion.

matrix completion, Toeplitz matrix, orthogonal, rank-one, mean value.