摘要
数字图像修复技术是对图像上信息缺损区域进行信息填充并使观察者无法察觉出图像曾经破损或已被修复.针对Criminisi算法随着修复的进行置信度迅速降为零从而造成修复顺序出现偏差的不足,提出一种改进的优先级函数.将优先级函数表示为置信度项和数据项的加权和,并将置信度项修改为指数函数形式以平滑其迅速降为零的趋势,从而使得修复顺序更加准确.此外,对优先级函数中的置信度项和数据项选取不同的权重因子可得到不同的修复图像供用户选择.实验结果表明,该算法取得了较好的修复效果.
Image inpainting is one of the technology, which can reach restore image information and is not easy to detected by people. This paper proposes a new priority function to improve the deficiency of inaccuracy the priority function of Cdminisi algorithm has. This method modifies the priority function from multiplication into addition and introduces exponential function to smooth the tendency that confidence term drop to zero dramatically, which can improve the priority computation precision. At the same time, this paper sets the weighting factor for a variety of restoration image for users to choose. Experimental results show that the proposed method achieves better inpalnting effect.
出处
《小型微型计算机系统》
CSCD
北大核心
2014年第12期2754-2758,共5页
Journal of Chinese Computer Systems
基金
广东省自然科学基金项目(S2011040004273)资助
关键词
图像修复
优先级函数
置信度
指数函数
权重因子
image inpainting
priority function
confidence term
exponential function
weighting factor