摘要
在分析了几种重要的基于滤波图像修复算法基础上,提出了一种改进的邻域滤波图像修复算法。为了减少平滑处理中的模糊,该算法根据待修复点与窗口中已知点的欧氏距离来决定相应位置的权值,然后采用邻域中已知像素的加权平均来估算待修复点的值。大量实验结果表明,改进算法修复的视觉效果和归一化均方误差(NMSE)的计算结果均明显优于先前的算法。
Several important filtering-based image inpainting algorithms are analyzed, an improved method for neighborhood filtering image inpainting is proposed. In order to reduce blurring in the smoothing process, the algorithm is based on the Euclidean distance between the damaged pixel and the useful surrounding pixels to determine the weights of the corresponding location, and then the guessed value of damaged pixel is to calculate the weighted average of its surrounding pixel colors. Experimental results show that the improved algorithm＇s visual effects and the result of normalized mean square error （NMSE） are much better than previous algorithms.
出处
《电视技术》
北大核心
2011年第11期14-16,35,共4页
Video Engineering
基金
国家自然科学基金青年基金项目(60802040)
四川省教育厅青年基金项目(07Zd1114)
关键词
邻域滤波
图像修复
加权平均
neighborhood filtering
image inpainting
weighted average
