一种基于非局部思想的改进图像降噪算法  

Improved image denoising algorithm based on non-local idea

作　　者：刘苒苒[1] 武小平[2] 韦超[2] 孔泽伦 

机构地区：[1]武汉理工大学汽车工程学院,武汉430079 [2]武汉大学计算机学院,武汉430072

出　　处：《计算机应用研究》2016年第4期1277-1280,共4页Application Research of Computers

基　　金：湖北省科技支撑计划资助项目(2014BAA149)

摘　　要：在基于稀疏和冗余字典的图像降噪算法的基础上,提出了一种基于非局部思想的改进图像降噪算法。与传统的基于稀疏表达的图像降噪算法K-SVD相比,提出的算法增加了一个相似块聚合的过程,使得学习的字典更小且更准确。利用自然图像包含很多的自相似,相似样本聚合学习出的字典比传统K-SVD算法能更准确更稀疏地表示样本。稀疏度的提高使得重建后的信号更加准确、适应性更好。实验证明提出的算法取得了更好的视觉效果。This paper presented an improved image denoising method based on sparse and redundant representations over trained dictionaries. With the traditional image denoising algorithm based on sparse expression compared to the K-SVD algorithm,it added a similar block polymerization process to build the smaller and more accurate learning dictionary. The novel idea behind the proposed approach was that natural images contained so many self-similarities that signal sparsity could be further promoted. This sparsity promotion made the learned dictionary more adaptive and accurate to restore the signal. The experimental results demonstrate that this proposed method provides better visual quality compared to the state-of-the-art methods.

关 键 词：图像降噪 稀疏表达 相似块聚合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP301.6[自动化与计算机技术—计算机科学与技术]