基于样本块的图像修补技术的改进

Modified exemplar-based image inpainting algorithm

改进了一种基于样本块的修补算法,提出了将原本的整体范围搜索改为在用户指定的范围内搜索,最大可能地保证了合适的结构信息被填入,保障修补后的图像在视觉上的连贯性,从而有效提高了大范围破损图像的修补效果.同时,将样本块窗口原有的大小从固定改进为根据图像的实际情况大小可变,使搜索速度得到较大提高.将该算法应用于遥感图像,取得了良好的视觉效果.

This paper changed the searching scope of the original exemplar based image inpainting from the whole image domain into a user-defined domain, which guarantees that the suitable structure is inpainted, and thereby efficiently improved inpainting results. Meanwhile, the origi- nal fixed size of the exemplar was also changed into an automatically adjusted size, which significantly improved the searching speed. This algorithm has been successfully applied to remotely sensed data.