摘要
针对K-奇异值分解(sigular value decomposition,SVD)算法存在的问题,结合结构聚类和字典学习,提出了一种基于非局部正则化稀疏表示的图像去噪算法。首先,利用非局部去噪的思想将结构相似的图像块聚类,每一类图像块单独进行字典学习,增强了字典的自适应性;其次,利用稀疏K-SVD替代传统的K-SVD进行类内字典学习,改善了字典的结构性;最后,引入稀疏系数误差正则项来修正稀疏系数以进一步改善图像的重构效果。实验结果表明,与传统的K-SVD算法相比,该算法能够有效地保持图像的结构信息,并且提升了去噪效果,同时,在不降低图像结构相似度的基础上,峰值信噪比很接近甚至部分好于目前先进的去噪算法。
For the existing problems of the K-sigular value decomposition (SVD) denoising method, a new denoising method based on non-local regularized sparse representation is proposed, which combines the structural clustering and dictionary learning. Firstly, image blocks that are similar in structure are clustered by using the idea of non-local denoising. It reinforces the adaptive ability of dictionary because each image block runs dictionary learning independently. Then, structured dictionaries within classes are learned through substituting the K-SVD by sparse K-SVD. Finally, in order to improve the effect of image recon- struction, the sparse coefficient error regularization is introduced to revise the sparse coefficient. Compared with the traditional K-SVD denoising algorithm, experiments show that the proposed method can protect the information of image structure effectively and promote the result of denoising greatly. Simultaneously, with out decreasing the structural similarity image measurement value, the peak signal to noise ratio value is very close and even better than the advanced denoising algorithm.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2013年第5期1104-1109,共6页
Systems Engineering and Electronics
基金
国家自然科学基金(61172127)
高等学校博士学科点专项科研基金(20113401110006)
安徽大学"211"工程学术创新团队基金(KJTD007A)
安徽省自然科学基金(1208085QF104)
安徽省高校优秀青年人才基金(2012SQRL017ZD)资助课题
关键词
非局部去噪
稀疏表示
结构聚类
字典学习
non-local denoising
sparse representation
structure clustering
dictionary learning