“…However, due to the ill-posed nature of deconvolution, the RL algorithm will amplify the noise with increasing iteration number. Thus, many regularizations such as Tikhonov regularization (TR) [9][10][11], shape priori [12], high-order statistics (HOS), Huber-Markov [13,14], and spatial regularization [15] have been incorporated. In recent years, total variation (TV) and its variants have been popular choices for the regularization terms to solve various blind deconvolution problems [16,17].…”