Atmospheric turbulence degrades images by causing non-uniform geometric deformations and distortions due to the random fluctuations of refractive index over air media. In this paper, a new approach based on hierarchical layer parsing is developed. Each frame from a set of temporal frames is decomposed into a collection of isoplanatic space-time patches. Thereafter, each of temporal patches is parsed hierarchically to separate layers corresponding to, low rank background, sparse turbulence errors, and moving objects, using low rank matrix decomposition. In order to remove redundant blur that remains after local hierarchical parsing blind de-convolution procedure is applied. Finally, the stable background component and the one or more moving object components are fused to reconstruct the turbulence mitigated content. © 2013 IEEE.