## Fan Yao*## |

Image size | [18] | [19] | [20] | Proposed | |||||
---|---|---|---|---|---|---|---|---|---|

Time (s) | PSNR (dB) | Time (s) | PSNR (dB) | Time (s) | PSNR (dB) | Time (s) | PSNR (dB) | ||

Fig. 5 | 682×570 | 76.8459 | 36.4581 | 75.42897 | 37.4529 | 63.3889 | 37.7219 | 81.4596 | 39.1562 |

Fig. 6 | 400×350 | 61.0259 | 36.4893 | 59.6489 | 37.2772 | 51.6972 | 37.4293 | 62.0348 | 39.1579 |

Fig. 7 | 432×306 | 60.8547 | 36.1457 | 59.4586 | 37.6578 | 53.2497 | 37.1689 | 61.0245 | 39.0389 |

Fig. 8 indicated the restoration effect on the damaged thangka image (Fig. 8(b)) of Haar wavelet (a support length of 1) and biorthogonal Bior4.4 wavelet (a support length of 9) with the wavelet decomposition level at 2 and 4, respectively.

Judging from Fig. 8, the restoration effect of Haar wavelet outperformed that of Bior4.4 wavelet, and the restoration effect of level-2 wavelet decomposition was better than that of level-4 wavelet decomposition. This was because when these data were set to zero during the restoration process in the damaged area, an artificial boundary was generated around the restoration area. When the wavelet was decomposed, the artificial boundary was caused by extension effect, which was related to the support length of base wavelet. The shorter the support length of base wavelet was, the smaller the elongation effect became, and the more accurate the restoration of the damaged area became. Therefore, in order to minimize the extension effect of the artificial boundary, the support length of base wavelet should be as short as possible. In addition, the wavelet decomposition series J was greater, and the low frequency energy was more concentrated. The low frequency subgraph restoration error would have great influence on the quality of the reconstructed image. The decomposition level J was preferably set at the level of 2 or 3, and they could respectively provide a subgraph of 1/4 or 1/8 of the original image. The decomposition level could basically meet the requirements of most restoration images.

In order to verify the effectiveness of this algorithm, the repair experiments of some actually damaged thangka images were carried out on computer. As shown in Figs. 9 and 10, Figs. 9(a) and 10(a) were actually damaged thangka images, and Figs. 9(b) and 10(b) were the images of damage marks. Figs. 9(c) and 10(c) were the inpainting results using literature [18]. Mismatches resulted in defects in the inpainting results. Figs. 9(d) and 10(d) were the restoration results using literature [19]. There were bulges and fractures in the restoration results. Figs. 9(d) and 10(e) were the restoration results using literature [20]. Although the damaged edges could be connected, the smooth area was blurry. Figs. 9(f) and 10(f) were the restoration results using the method presented in this paper. The results based on this algorithm displayed original visual effect. This algorithm enjoyed good restoration effect in both smooth and edge parts.

Because Figs. 9 and 10 were both actually damaged images, their PSNR values could not be calculated. Objective judgment could only rely on the inpainting time of algorithms, and the repair time of the three algorithms was listed in Table 2.

It could be seen from Table 2 that the restoration time of this method was slightly longer than that of other algorithms. The reason was that when the high frequency subgraph was restored, the consumed time of the algorithm increased due to increased curve fitting, edge factor and improved search method.

When some damaged edges are restored, it is difficult to take into account the structural integrity of an image and good visual effect on the basis of traditional algorithms. In this paper, an inpainting algorithm combining wavelet transform with structure constraint has been proposed for thangka images. Thangka images were decomposed into subgraphs with different resolutions and components by wavelet transform, and then the edge contour information was extracted and restored. Moreover, the coefficient relations of different components were organically combined with texture synthesis, and the edge contour information was extracted and repaired to constrain the repair of the high frequency subgraph structure. By introducing edge change factor and improving the matching block search method, the repair methods of layering and classification were adopted to enhance the ability to restore edges and textures. Experimental results showed that this algorithm could effectively restore damaged images with strong edges and rich texture, and had a better restoration effect which was more in line with the visual effect of human eyes. However, the algorithm took longer to restore images, and it was not ideal to restore large texture damaged images. Therefore, how to improve the restoration accuracy of low frequency subgraphs is still worth further discussion.

- 1 S. T. Pohlmann, E. F. Harkness, C. J. Taylor, S. M. Astley, "Evaluation of Kinect 3D sensor for healthcare imaging,"
*Journal of Medical and Biological Engineering*, vol. 36, no. 6, pp. 857-870, 2016.custom:[[[-]]] - 2 M. Imtiyaz, A. Kumar, S. Sreenivasulu, "Inpainting an image based on enhanced resolution,"
*Journal of Computer Technology & Applications*, vol. 6, no. 1, pp. 23-26, 2015.custom:[[[-]]] - 3 S. Feng, R. Murray-Smith, A. Ramsay, "Position stabilisation and lag reduction with Gaussian processes in sensor fusion system for user performance improvement,"
*International Journal of Machine Learning and Cybernetics*, vol. 8, no. 4, pp. 1167-1184, 2017.doi:[[[10.1007/s13042-015-0488-5]]] - 4 F. Li, T. Zeng, "A new algorithm framework for image inpainting in transform domain,"
*SIAM Journal on Imaging Sciences*, vol. 9, no. 1, pp. 24-51, 2016.doi:[[[10.1137/15M1015169]]] - 5 F. Chen, T. Hu, L. Zuo, Z. Peng, G. Jiang, M. Yu, "Depth map inpainting via sparse distortion model,"
*Digital Signal Processing*, vol. 58, pp. 93-101, 2016.doi:[[[10.1016/j.dsp.2016.07.019]]] - 6 W. Chen, H. Yue, J. Wang, X. Wu, "An improved edge detection algorithm for depth map inpainting,"
*Optics and Lasers in Engineering*, vol. 55, pp. 69-77, 2014.custom:[[[-]]] - 7 F. Wang, D. Liang, N. Wang, Z. Cheng, J. Tang, "An new method for image inpainting using wavelets," in
*Proceedings of 2011 International Conference on Multimedia Technology*, Hangzhou, China, 2011;pp. 201-204. custom:[[[-]]] - 8 H. Zhang, S. Dai, "Image inpainting based on wavelet decomposition,"
*Procedia Engineering*, vol. 29, pp. 3674-3678, 2012.custom:[[[-]]] - 9 A. S. Deshmukh, P. Mukherji, "Image inpainting using multiresolution wavelet transform analysis," in
*Proceedings of 2012 International Conference on Communication*, Information & Computing Technology (ICCICT), Mumbai, India, 2012;pp. 1-6. custom:[[[-]]] - 10 K. He, R. Liang, T. Zhang, "Fast texture image completion algorithm based on dependencies between wavelet coefficients,"
*Journal of Tianjin University*, vol. 43, no. 12, pp. 1093-1097, 2010.custom:[[[-]]] - 11 Z. Xiao, W. Zhang, "Wavelet-domain fast inpainting algorithm for texture image,"
*Chinese Journal of Scientific Instrument*, vol. 29, no. 7, pp. 1422-1425, 2008.custom:[[[-]]] - 12 M. Xiao, G. Li, Y. Tan, J. Qin, "Image completion using similarity analysis and transformation,"
*International Journal of Multimedia and Ubiquitous Engineering*, vol. 10, no. 4, pp. 193-204, 2015.custom:[[[-]]] - 13 A. Telea, "An image inpainting technique based on the fast marching method,"
*Journal of Graphics Tools*, vol. 9, no. 1, pp. 23-34, 2004.doi:[[[10.1080/10867651.2004.10487596]]] - 14 P. Buyssens, M. Daisy, D. Tschumperle, O. Lezoray, "Exemplar-based inpainting: technical review and new heuristics for better geometric reconstructions,"
*IEEE Transactions on Image Processing*, vol. 24, no. 6, pp. 1809-1824, 2015.doi:[[[10.1109/TIP.2015.2411437]]] - 15 L. Cai, T. Kim, "Context-driven hybrid image inpainting,"
*IET Image Processing*, vol. 9, no. 10, pp. 866-873, 2015.doi:[[[10.1049/iet-ipr.2015.0184]]] - 16 J. Patel, T. K. Sarode, "Exemplar based image inpainting with reduced search region,"
*International Journal of Computer Applications*, vol. 92, no. 12, pp. 27-33, 2014.custom:[[[-]]] - 17 L. Tang, Y. Tan, Z. Fang, C. Xiang, S. Chen, "An improved Criminisi image inpainting algorithm based on structure component and information entropy,"
*Journal of Optoelectronics· Laser*, vol. 28, no. 1, pp. 108-116, 2017.custom:[[[-]]] - 18 H. M. Patel, H. L. Desai, "A review on design, implementation and performance analysis of the image inpainting technique based on TV model,"
*International Journal of Engineering Development and Research (IJEDR)*, vol. 2, no. 1, pp. 191-195, 2014.custom:[[[-]]] - 19 H. Zhang, S. Dai, "Image inpainting based on wavelet decomposition,"
*Procedia Engineering*, vol. 29, pp. 3674-3678, 2012.custom:[[[-]]] - 20 B. Wang, L. Hu, J. Cao, R. Xue, G. Liu, "Image restoration based on sparse-optimal strategy in wavelet domain,"
*Acta Electronica Sinica*, vol. 44, no. 3, pp. 600-606, 2016.custom:[[[-]]]