Research on texture image inpainting of jacquard fabric based on non-single vision

For the texture image inpainting of jacquard fabric with a large damaged region but complete data information around it, traditional exemplar-based image inpainting algorithms are not only have rapid falling traditional confidence value, but also the matching precision is greatly limited due to the lack of robustness, which results in a wrong guided direction and an unsatisfactory inpainting effect. To solve the above-mentioned problems, a novel digital image acquisition system for damaged fabric was designed and one set of texture image inpainting algorithms of jacquard fabric based on non-single vision was developed. In this image inpainting method, the images to be inpainted and target images were obtained through the non-single-vision imaging method. By using the image registration and Poisson image blending technology, the damaged region was filled by block matching first and then patches from the target image to the data missing region were copied. The experimental results clearly showed that the proposed method achieves better results in visual appearance and inpainting quality, compared with traditional approaches for inpainting. In addition, this method shows broad potential application prospects in the virtual restoration of ancient fabrics using image inpainting technology.