摘要
针对经典的Criminisi图像修复算法中出现与周围邻域像素不协调的块状效应,在深入研究边界像素点权值的基础上,提出了一种改进样本块图像修复算法。该算法引进灰度熵来改进优先权值计算方法 ,进而建立新的优先级修复序列,并在寻找最佳匹配块时考虑块修复后与周围邻域的连贯性。仿真实验表明,该算法在一定程度上弥补了经典Criminisi图像修复算法的不足。相比传统样本块图像修复算法,运行时间有所缩短,修复效果好、画面更和谐、自然,符合人类审美观。
A new improved image inpainting algorithm is proposed through deeply study the weight value of the pixels, because classic Criminisi image inpainting algorithm has the block effect of not harmonious with neighbors. The new algorithm mainly improves the priority of the block by bringing in grey entropy and thinks about being consistent with the surrounding neighborhood when finding the most similar image block. Many simulation experiments results show that the new algorithm makes up for the deficiency in inpainting effect, shorter running time, image harmony according to human aesthetic appreciation comparing to traditional exemplar image inpainting algorithm.
出处
《自动化技术与应用》
2017年第3期59-63,共5页
Techniques of Automation and Applications
关键词
图像修复
像素权值
灰度熵
样本块
image inpainting
pixel priority
grey entropy
exemplar
