Restoration Method Based on Low-Rank Decomposition for Video under Turbulence

Atmospheric turbulence can cause the image degraded with blur and distortion. These make the image reconstruction big challenges. In order to correct pixel distortion and reduce blur in the turbulence degraded videos, a novel video restoration algorithm based on low-rank decomposition is proposed. First it analyzes the characteristics of atmospheric turbulence degradation. Next, according to the low-rank decomposition, it decomposes the turbulence sequence into two components: the low-rank stabilized background and the sparse turbulence pixel offsets. The object function of decomposition, which is a combination of the matrix nuclear norm and Frobenius norm, is optimized by inexact augmented Lagrange multiplier. Finally, the stabilized frames can be restored in a frame-by-frame basis. Experimental results show that the effectiveness of the proposed method. It can effectively alleviate pixel distortions and blur, improve visual quality significantly.