Fig. 1. Description of Criminisi algorithm

Fig. 2. Algorithm block diagram of textile cultural relic image restoration

Fig. 3. Boundary tracking results after denoising. (a) Raw wool fabric image; (b) boundary tracking images without denoising; (c) boundary tracking image processed by adaptive filtering; (d) boundary tracking image processed by proposed algorithm

Fig. 4. Schematic diagram of mask making process. (a) Original image; (b) damaged area; (c) mask map; (d) image to be repaired

Fig. 5. The result of K-means color segmentation

Fig. 6. Color dispersion of different pixel blocks. (a) Sample 1,$\sigma (p)=\mathrm{0.3398}$; (b) Sample 2,$\sigma (p)$=0.5350

Fig. 7. Comparison of inpainting results of damaged textile cultural relics images by four algorithms. (a) Image of damaged textile artifacts; (b) mask map; (c) inpainting image by Criminisi algorithm;(d) inpainting image by reference [16] algorithm; (e) inpainting image by reference [22] algorithm; (f) inpainting image by proposed algorithm

Fig. 8. Comparison of repairing effect of artificial fictitious damaged textile image. (a) Image of original textile; (b) artificial virtual damage image; (c) inpainting image by Criminisi algorithm;(d) inpainting image by reference [16] algorithm; (e) inpainting image by reference [22] algorithm; (f) inpainting image by proposed algorithm

Fig. 9. Comparison of inpainting effects. (a) Original image; (b) damaged image; (c) inpainting image by reference [16] algorithm; (d) inpainting image by reference [22] algorithm; (e) inpainting image by proposed algorithm

Table 1. Comparison of quality evaluation parameters (test 1)

Table 2. Comparison of quality evaluation parameters (test 2)