Fig. 1. Structure of light field imaging system

Fig. 2. Light field camera parameters and images. (a) Parameters of light field camera; (b) remote sensing image in the UMLUD dataset; (c) raw image of light field; (d) light field sub-aperture image

Fig. 3. Correspondence between refocusing image of light field and Laplace operator

Fig. 4. Refocus processed images sequence

Fig. 5. Dataset image and image taken by real cameras. (a) Image taken by real camera; (b) remote sensing image of the UMLUD dataset

Fig. 6. The images processed by each deblurring algorithm. (a) Light field real blurred image; (b) real image processed by DCP;(c) real image processed by L0; (d) real image processed by PMP; (e) real image processed by NLC; (f) real image processed by proposed algorithm; (g) UMLUD remote sensing blurred image; (h) remote sensing image processed by DCP; (i) remote sensing image processed by L0; (j) remote sensing image processed by PMP; (k) remote sensing image processed by NLC; (l) remote sensing image processed by proposed algorithm

Fig. 7. Image evaluation indicators of each algorithm on Levin dataset. (a) Error ratio curves; (b) PSNR; (c) SSIM; (d) Laplace

Fig. 8. Algorithm convergence test. (a) Convergence curve of proposed algorithm on the energy equation; (b) convergence curve of proposed algorithm on kernel similarity

Table 1. Image evaluation index after deblurring for real images and UMLUD remote sensing images

Table 2. The running time of each algorithm for processing images of different size