摘要
提出一种改进的基于K-SVD字典的图像修复算法.该算法基于稀疏表示,利用待修复图像内的有效信息,以不重叠像素的方式提取图像块,采用模糊C均值聚类算法对图像块进行聚类,并使用K-SVD算法分别对各类图像块进行训练,得到与各类图像块相适应的字典,重建图像块,修复受损图像.实验结果表明,该算法能提高图像的修复质量和图像的峰值性噪比,且均方根误差较小.
This dictionary. The paper proposed algorithm based an on improved inpainting algorithm which was based on K-SVD sparse representation used valid image information, extracted image blocks with no overlaps, adopted fuzzy C-means clustering algorithm to cluster image blocks, and used K-SVD algorithm to train dictionary for each class, obtained dictionary which was adapted to each class, reconstructed image blocks, restored this algorithm repaired signal-to-noise ratio. damaged image well, had a damaged image. Experimental results showed that smaller root mean square error, and improved peak
出处
《安徽大学学报（自然科学版）》
CAS
北大核心
2013年第3期69-74,共6页
Journal of Anhui University(Natural Science Edition)
基金
安徽省教育厅自然科学基金重点资助项目(KJ2009A60)
安徽大学博士科研启动基金资助项目(33190049)
安徽大学"211工程"学术创新团队基金资助项目(KJTD007A)
关键词
稀疏表示
图像修复
K-SVD
训练字典
模糊C均值聚类
sparse representation
image inpainting
K-SVD
training dictionary
fuzzy C-means clustering
