摘要
针对传统Criminisi算法在修复图像时,存在置信度项快速趋向于零和匹配样本块的搜索策略复杂度高的问题,论文提出一种结合阈值分界和区域划分的图像修复算法。首先通过阈值分界的方式改进优先权函数,其次引入Jaccard距离作为匹配准则的约束条件,最后通过区域分割的方式优化搜索策略。实验结果分析表明,新算法可以较好地克服纹理延伸的问题,获得更好的峰值信噪比和结构相似性。
In order to avoid the problems that the confidence item quickly tends to zero and the search strategy of matching sample blocks is complicated when Criminisi algorithm is repairing the image,this paper proposes an inpainting algorithm based on threshold demarcation and region division.Firstly,the priority function is improved by the method of threshold demarcation.Second-ly,the Jaccard distance is introduced as the constraint condition of the matching criterion.Finally the search strategy is optimized by the method of region segmentation.Experimental results show the improved algorithm can better overcome the problem of texture extension,and obtain better peak signal-to-noise ratio and structural similarity.
作者
周先春
陈璟
张婕
殷新鹏
郭亮可
ZHOU Xianchun;CHEN Jing;ZHANG Jie;YIN Xinpeng;GUO Liangke(College of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing 210044)
出处
《计算机与数字工程》
2023年第6期1393-1397,1415,共6页
Computer & Digital Engineering
