激光与光电子学进展, 2020, 57 (20): 201020, 网络出版: 2020-10-17
基于序贯相似性和布谷鸟寻优的敦煌壁画修复算法 下载: 904次
Dunhuang Mural Inpainting Algorithm Based on Sequential Similarity Detection and Cuckoo Optimization
图像处理 壁画修复 序贯相似性检测 布谷鸟寻优算法 Criminisi算法 image processing mural inpainting sequential similarity detection cuckoo optimization algorithm Criminisi algorithm
摘要
针对Criminisi算法在修复敦煌壁画时易出现错误填充、修复效率较低等问题,提出了一种基于序贯相似性和布谷鸟寻优算法结合的敦煌壁画修复方法。首先采用P-Laplace算子重新定义数据项,改进了优先权计算方法,避免了优先权频繁趋于0的问题;其次引入动态阈值序贯相似性检测算法进行匹配块的搜索,提高了壁画修复效率;为了使匹配块选择更加合理,再利用布谷鸟寻优算法确定最佳匹配块;最后通过迭代更新完成壁画修复。通过对敦煌壁画的修复实验表明,本文方法相比于同类比较算法,取得了较好的主客观修复效果,并且修复效率也得到了进一步提升。
Abstract
In this paper, we proposed the Dunhuang inpainting mural restoration algorithm based on the combination of sequential similarity detection algorithm and cuckoo search algorithm to improve the incorrect filling problem of the Criminisi algorithm and low efficiency in Dunhuang murals restoration. First, we improved the priority calculation formula with the method of redefining data items using a P-Laplace operator to eradicate the priority calculation tends to zero. Second, we improved the efficiency of mural restoration using the sequential similarity detection algorithm based on the dynamic threshold for searching matching blocks. To make the matching block more reasonable, we used a cuckoo optimization algorithm to determine the best matching block. Finally, mural restoration was completed through iterative updates. The results of restoration experiments on Dunhuang mural inpainting show that compared with similar comparison algorithms, the proposed algorithm in this paper achieves better subjective and objective restoration effects, and improves the restoration efficiency.
陈永, 陈锦, 艾亚鹏, 陶美风. 基于序贯相似性和布谷鸟寻优的敦煌壁画修复算法[J]. 激光与光电子学进展, 2020, 57(20): 201020. Yong Chen, Jin Chen, Yapeng Ai, Meifeng Tao. Dunhuang Mural Inpainting Algorithm Based on Sequential Similarity Detection and Cuckoo Optimization[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201020.