基于纹理结构引导自适应的改进Criminisi算法

针对缺损图像修复时容易产生纹理紊乱、边界残缺等问题,基于Criminisi算法提出一种纹理结构引导的自适应图像修复算法。首先对决定合成顺序的优先级进行改进,在数据项中加入结构张量,使图像修复从结构区域向无结构区域填充;其次根据原图像区域纹理结构信息自适应地改变模板块的尺寸,这在一定程度上避免了纹理块过小或过大带来的弊端,从而使合成效果更为自然。实验结果表明,本文提出的改进算法不仅保证了修复图像结构信息的合理填充,还较好地保持了修复边界的完整性,图像修复后具有较佳视觉效果。

Improved Criminisi algorithm based on texture structure guided self-adaptation

For the problems of texture disorder and boundary defect in restored images, this paper proposes a texture structure guided self-adaptive image inpainting approach based on Criminisi algorithm. Firstly, the priority which determines the synthetic sequence is improved by adding structure tensor in data items, so the demaged image is filled from the structural region to the nonstructural ones. Secondly, according to texture information in the original image, the size of each template block is self-adaptively changed, which avoids the disadvantage of improper texture patches to a certain extent, and makes the synthetic effect more natural. Experimental results show that the improved algorithm not only guarantees the reasonable filling of the structural information in the images but also maintains the integrity of the repaired boundaries, and the restored images have better visual effect.