摘要
使用块拼贴的基于样本的图像修复算法运行时间主要取决于最佳样本块匹配步骤的执行效率。目前算法普遍采用全局搜索获取样本块,逐一与待修复块进行相似性比对,修复质量和修复效率依赖于采样区域的范围大小。为提高计算效率,提出一种基于局部平均灰度熵的图像修复算法,在每次迭代中根据待修复块邻域窗的平均灰度熵自适应确定采样区域范围。实验结果证明,所提算法相较经典Criminisi修复算法提高了修复质量,且大大提高了修复效率。
The operation time of sample-based image inpainting algorithms using block collage mainly depends on execution efficiency of the best sample blocks matching procedure. Currently these algorithms commonly adopt global search to obtain sample blocks for similarity comparison with the blocks to be inpainted one by one, and the range size of the sampling region has great influence on the inpainting quality and efficiency. For improving computation efficiency, we propose a novel image inpainting algorithm which is based on local average gray entropy, in iteration it determines adaptively the size of sampling region according to the average gray entropy of the neighbouring window of the block to be inpainted every time. Experimental results show that the proposed algorithm improves the inpainting quality compared with classic Criminisi inpainting algorithm, besides, it also greatly raises the inpainting efficiency.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第10期206-208,223,共4页
Computer Applications and Software
基金
国家自然科学基金项目(60903186
61401281)
上海市自然科学基金(14ZR1440700)
上海市高校青年教师培养资助计划(ZZyyy13022)
上海应用技术学院引进人才基金(YJ201310)
关键词
图像修复
图像补全
纹理合成
局部平均灰度熵
Image inpainting
Image completion
Texture synthesis
Local average gray entropy
