Fig. 1. Schematic of two-dimensional Gaussian model

Fig. 2. Defocusing blurred kernel (upper) and its one-dimensional profile (down). (a)(d) Disk model; (b)(e) Airy spot; (c)(f) convolution result

Fig. 3. Estimated actual blurred kernel (right) and its x-axis one-dimensional profile (left). (a)(b) First; (c)(d) second; (e)(f) third

Fig. 4. VPW-function-fitted curves under different parameters

Fig. 5. Fitting results of actual blurred kernel based on different models

Fig. 6. Evaluation results of different kernels. (a) RMSE values; (b) R-square values

Fig. 7. Blurred step edges

Fig. 8. Detection results of edge points

Fig. 9. Detection results of ESF curves. (a)(b) Before interpolation; (c)(d) after interpolation

Fig. 10. R-L image restoration method based on VPW fitting PSF

Fig. 11. Original images and composite blurred kernels

Fig. 12. Restoration results by method in Ref. [20]

Fig. 13. Restoration results by method in Ref. [16]

Fig. 14. Restoration results by method in Ref. [4]

Fig. 15. Restoration results by proposed method

Fig. 16. RMSE fitted by different kernel models

Fig. 17. Restoration results of house image. (a) Actual blurred image; (b) restoration result in Ref. [20]; (c) restoration result in Ref. [16]; (d) restoration result in Ref. [4]; (e) restoration result here

Fig. 18. Restoration results of track image. (a) Actual blurred image; (b) restoration result in Ref. [20]; (c) restoration result in Ref. [16]; (d) restoration result in Ref. [4]; (e) restoration result here

Fig. 19. Restoration results of harbor image. (a) Actual blurred image; (b) restoration result in Ref. [20]; (c) restoration result in Ref. [16]; (d) restoration result in Ref. [4]; (e) restoration result here