Author Affiliations
School of Electronics and Information Engineering, Lanzhou Jiaotong University, Lanzhou, Gansu 730070, Chinashow less
Fig. 1. Principle of the Criminisi algorithm
Fig. 2. Images before and after edge fitting. (a) Image with edge missing; (b) image with edge fitting
Fig. 3. Fitting process of the second-order Bézier curve
Fig. 4. Schematic diagram of the second-order Bézier curve
Fig. 5. Schematic diagram of Bézier curve fitting. (a) Curve before fitting; (b) curve after fitting
Fig. 6. Edge fitting experiment of damaged murals. (a) Original image; (b) mask image; (c) manually selected control point 1; (d) manually selected control point 2; (e) our algorithm
Fig. 7. Process of fusion to generate image to be repaired
Fig. 8. Repair results of different algorithms. (a) Original mural; (b) mask image; (c) Criminisi algorithm; (d) our algorithm
Fig. 9. Repair result of the real mural. (a) Original mural; (b) mask image; (c) Criminisi algorithm; (d) our algorithm; (e) local repair result of Criminisi algorithm; (f) local repair result of our algorithm
Fig. 10. Repair results of different algorithms. (a) Original image; (b) mark image; (c) XTh = 2000; (d) XTh = 1000; (e) XTh = 500; (f) XTh = 300; (g) XTh = 200; (h) XTh =100; (i) XTh =50; (j) our algorithm
Fig. 11. Repair results of artificially damaged murals by different algorithms. (a) Original mural; (b) artificially damaged mask image; (c) Criminisi algorithm; (d) algorithm of Ref. [8]; (e) algorithm of Ref. [10]; (f) our algorithm
Fig. 12. Repair result of the real broken mural. (a) Original mural; (b) mask image; (c) Criminisi algorithm; (d) algorithm of Ref. [8]; (e) algorithm of Ref. [10]; (f) our algorithm
Mural | Criminisi | Ref. [8] | Ref. [10] | Ours |
---|
1 | 34.70 | 37.04 | 37.65 | 44.93 | 2 | 37.31 | 32.52 | 34.98 | 44.93 | 3 | 32.95 | 33.65 | 32.52 | 38.08 | 4 | 41.10 | 32.76 | 29.75 | 41.86 | 5 | 29.34 | 29.11 | 30.02 | 37.05 |
|
Table 1. PSNR of image repaired by different algorithms unit: dB
Mural | Criminisi | Ref. [8] | Ref. [10] | Ours |
---|
1 | 0.9870 | 0.9882 | 0.9877 | 0.9929 | 2 | 0.9803 | 0.9777 | 0.9804 | 0.9830 | 3 | 0.9912 | 0.9912 | 0.9860 | 0.9947 | 4 | 0.9804 | 0.9811 | 0.9762 | 0.9849 | 5 | 0.9785 | 0.9816 | 0.9796 | 0.9905 |
|
Table 2. SSIM for image restoration with different algorithms