摘要
针对目前图像填补中存在的很难自动分析和提取前景、背景,然后合理地恢复图像等问题,提出一种利用纹理规则性分析和纹理合成技术进行有效图像填补的方法.为了实现自动化的提取,将前景视为近似规则纹理的场景区域,首先引入纹理规则性分析方法,利用近似规则纹理具有强烈的信号规则性的特点自动有效地提取前景信息;然后通过合成优先级的思想将纹理合成技术和图像修补技术结合起来,提出一个合理的背景填补方法以恢复图像.实验结果证明,该方法可以自动有效地提取具有近似规则纹理特征的前景或背景内容,并合理地恢复图像.
In order to extract foreground contents within an image automatically and reconstruct the background in a visually plausible way,this paper proposes a method for image completion based on texture regularity analysis and texture synthesis technology.In view of the difficulty of automation,the method focuses on a special kind of foreground contents called near-regular texture,NRT.Firstly,a texture regularity analysis is adopted so that the foreground contents in strong regularity of NRT can be automatically extracted effectively.Secondly,by introducing a priority calculation,a combined framework of texture synthesis and image inpainting is applied to reconstructing the background properly.Experiments show that this method could be employed to automatically extract foreground or background with the feature of near regular textures and recover the image in good visual quality.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2012年第2期183-190,共8页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金重点项目(60833007)
澳门大学研究基金与研究生奖学金
关键词
图像填补
纹理合成
纹理规则性
近似规则纹理
优先级计算
图像修补
image completion
texture synthesis
texture regularity
near-regular texture
priority calculation
image inpainting
