基于结构因子和颜色聚类的墓葬图像修复算法

墓葬壁画图像颜色丰富,一旦破损会丢失大量结构信息。传统算法修复此类图像时,没有考虑图像强结构信息的优先精确修复,造成修复区域的过延伸和不连贯。针对上述问题,提出了一种基于结构因子和颜色聚类的墓葬图像修复算法。算法首先在待修复块优先级计算中加入结构因子项;其次,通过颜色FCM聚类算法划分区域进行相似块精确搜索;最后,根据均值像素差平方和(ASSD)与设定阈值的大小关系,自适应地对修复块尺寸进行调整以实现复杂结构区域的精确匹配。实验结果表明,所提方法对北齐墓葬壁画图像大面积缺损有很好的修复效果,与Criminisi算法比较,在结构相似度(SSIM)上至少提升5.68%。

Restoring Algorithm of The Tomb Image Based on Structure Factor and Color Glustering

Image of the tomb mural has rich color, once destroyed it will lose a lot of structural information. Traditional algorithms restoring such image haven"t considered the priority of accurate restoration of strong structural information which results in repair area over extended and incoherent. To solve this problem, an image restoration algorithm based on structure factor and color clustering is proposed. Firstly, structure factor items are added in the priority calculation of repairing block. Secondly, search for similar block by color FCM clustering algorithm. Finally, depending on the size relationship between Average sum of Squared Differences (ASSD) and the set threshold, adaptively adjust the size of the repair block to achieve accurate matching of complex structure of the region. In this paper, experimental results show that a method of image restoration for the Northern Qi Dynasty tomb murals has a good result. Compared with Criminisi algorithm, structural similarity (SSIM) upgrades at least 5.68%.