Motion blur parameters estimation of uncooled infrared images

Motion blur widely exists in uncooled infrared images. Because of the obvious cross-shaped bright lines and a large amount of noise interference in the uncooled infrared image spectrum, it is difficult to estimate the point spread function (PSF) parameters. Aiming at the problem that the existing PSF estimation methods have strong parameter dependence and are greatly affected by interference in small blurred scale (the effective estimation range is limited), which leads to low accuracy of parameter estimation, this paper proposes a motion blurred PSF parameter estimation method based on spectrum clipping and structured forests edge detection. In order to avoid the influence of circular fringes and ensure the efficiency and accuracy of the algorithm, the obvious part of the middle fringes in the spectrum is cut out for the estimation of PSF parameters. Gauss low-pass filter and edge detection based on structured forests extract the fringe edge of the spectrum, overcoming the interference of cross-shaped bright lines and noise, and improving the accuracy of estimation results. Experimental results show that the proposed method is suitable for both visible and uncooled infrared motion blurred images. The PSF parameter estimation results have high accuracy, and our method has small parameter dependence and strong robustness.