Abstract: The traditional image inpainting algorithm which is based on partial differential equations requires a large number of iterations, so it consumes a lot of time and is more complicated. To solve this problem, a non-iterative adaptive image inpainting algorithm in the wavelet domain is presented. The damaged image is decomposed by wavelet transformation. According to both the area to be inpainted and the pixels around it, a mask is chosen adaptively. And the pixels can be selected according to its direction, so that the inpainting process is strictly performed according to the direction of isophote. The inpainted image is reconstructed by wavelet reconstruction. The experimental results show that the proposed algorithm not just can shorten time, and can reconstruct both texture detail and structure information of the damaged image with better performance.

Key words: image inpainting, wavelet decomposition, adaptive, direction filter, wavelet reconstruction

摘要： 传统的基于偏微分方程的图像修复算法需要大量迭代，修复所耗时间较长，复杂度高。针对这一问题，提出了一种小波域的非迭代自适应图像修复算法。该算法对破损图像进行小波分解，找到待修复区域，根据待修复区域及其邻域像素值自适应选择修复模板大小，对修复模板内的像素值进行方向筛选，使修复过程严格按照等照度线方向行进，对修复后的图像进行小波重构。实验结果表明，该方法显著地缩短了修复时间，且对于图像的纹理细节、结构信息都达到了更好的修复效果。

关键词: 图像修复, 小波分解, 自适应, 方向筛选, 小波重构