Exemplar-Based Image Inpainting using Automatic Patch Optimization
Xuehui Bi, Huaming Liu, Guanming Lu, Jian Wei, Chao Yan · 2018
Image inpainting has a wide range of applications in image processing. Exemplar-based technique can inpaint texture and structure regions simultaneously. However, the size of patch influences the result of inpainting. So far, there is no easy way to automatically determine the size of patch. Structure tensor can be used to determine priority, because it is able to determine the properties of a local area. In the paper, the size of patch is computed by structure tensor when adopting exemplar-based technique. To reduce blockiness, boundary constraints is added for searching for similar patches. The experiments show our proposed method can inpaint texture, flat, structure images. Adding boundary constraints can eliminate blockiness to a certain extent.