Fig. 1. General image degradation model

Fig. 2. Local minimum pixel intensity histograms of clear images and corresponding blurred remote sensing images, the size of local blocks is 45×45. (a) Local minimum pixel intensity histogram of clear and blurred optical remote sensing images; (b) local minimum pixel intensity histogram of clear and blurred infrared remote sensing images

Fig. 3. Local maxmum pixel intensity histograms of clear images and corresponding blurred remote sensing images, the size of local blocks is 45×45. (a) Local maxmum pixel intensity histogram of clear and blurred optical remote sensing images; (b) local maxmum pixel intensity histogram of clear and blurred infrared remote sensing images

Fig. 4. Flowchart of blind deblurring algorithm

Fig. 5. Four blur kernels used in this work. (a) 17×17 kernel; (b) 19×19 kernel; (c) 21×21 kernel; (d) 23×23 kernel

Fig. 6. Results of restoration of visible blurred remote sensing images by the proposed algorithm. (a) (c) (e) (g) Blurred images and real blur kernels; (b) (d) (f) (h) deblurring results and estimated blur kernels

Fig. 7. Comparison of image details after restoration by different methods. (a) Ground-truth images; (b) deblurring results of Krishnan-11; (c) deblurring results of Bai-18; (d) deblurring results of proposed method

Fig. 8. Quantitative evaluation results of 5 blind restoration algorithms for optical remote sensing images. (a) Average PSNR; (b) average SSIM

Fig. 9. Quantitative evaluation results of 5 blind restoration algorithms for infrared remote sensing images. (a) Average PSNR; (b) average SSIM

Fig. 10. Running time of blind restoration algorithms. (a) Running time for processing dataset of optical remote sensing images; (b) running time for processing dataset of infrared remote sensing images