文献类型：期刊文献

中文题名：一种改进的基于Criminisi算法的目标移除方法

英文题名：An Improved Method of Object Removal Based on Criminisi Method

作者：池悦[1];何宁[2];张琪[1];赵珊珊[1]

第一作者：池悦

机构：[1]北京联合大学北京市信息服务工程重点实验室;[2]北京联合大学信息学院

第一机构：北京联合大学北京市信息服务工程重点实验室

年份：2017

卷号：31

期号：1

起止页码：67-74

中文期刊名：北京联合大学学报

外文期刊名：Journal of Beijing Union University

基金：国家自然科学基金项目资助(61572077)

语种：中文

中文关键词：纹理合成;图像修复;优先权;局部匹配

外文关键词：Texture synthesis; Image restoration ; Priority; Local matching

摘要：由Criminisi算法设想并提出的基于样本块图像修复方法,不但能够修复大面积破损区域,而且可以用于移除图像中的目标物。针对其在计算修复块优先级和搜索最佳匹配块时存在的一些不足,提出了一种改进的基于样本块的目标移除方法。在图片预处理中首先进行膨胀腐蚀处理;其次优先级计算中将置信度计算引入指数,从而得到更加精确的优先权;再次运用局部搜索方法,大大提高了搜索速度和准确性;最后将整幅图再进行拉普拉斯锐化。实验结果表明,修复后的图像效果良好,峰值信噪比大大提高,并且提升了算法效率。
The block-based image restoration method is proposed by Criminisi algorithm. It not only can repair the large-area damaged image, but also can be used to remove the objects in the image, tn this paper, we propose an improved method of object removal based on sample blocks for calculating the block priority and searching for the best matching block. In the image preprocessing, the expansive etching process is firstly carried out. Secondly, the confidence calculation is introduced into the index to get the more accurate priority. Thirdly, the local search method is used again, which greatly improves the searching speed and accuracy. Finally, a Laplaeian sharpening is performed. The experimental results show that the reconstructed image has good effect and the peak signal-to- noise ratio is greatly improved, and the efficiency of the algorithm is improved.