Multi-frame image restoration based on a new degradation model of hazy and blurred image
Mengdi Wang, Shuyin Tao, Wende Dong, Qiong Wang · 2019
The quality of outdoor images is significantly influenced by atmospheric environment and the vibration of imaging system. In this paper, we propose a novel model to describe the process of image degradation due to the above two factors, and a blind image restoration method based on this model. We derive the degradation model from the hypothesis that the process of atmospheric degradation and the process of vibration causing motion blur are independent. The contrast and brightness of images is decreased due to atmospheric absorption and scattering. Besides, multiple scattering and camera vibration lead to image blurring. We estimate the atmospheric light and transmittance, therefore the image restoration problem is converted to a blind deconvolution problem. A deblurring method of multi-frame image is proposed under the Bayesian posterior probability model, where the latent images are regularized by a natural image gradient prior and the point spread functions are regularized by an l1-norm based prior. An alternating minimization approach is used to solve the optimization problem. Experimental results of synthetic images and real world images demonstrate that our model can well describe the degradation process and the proposed method can recover images effectively.