Abstract: Reconstructing the full covariance matrix from incomplete measurements of a non-redundant sparse linear array (NRA) is crucial for high-resolution direction-of-arrival (DoA) estimation.