基于结构相似度衡量的图像超分辨率重建

摘要

图像超分辨率一直都是图像处理领域的研究热点，提出了把图像质量评价方法用于图像超分辨率的方法。把结构相似性度量用于块匹配，搜索图像块的相似块，利用找到的相似块的加权平均重构高分辨率图像。在低分辨率图像上进行了有噪声，无噪声情况下的两种实验，实验结果证明本文提出的方法在相同实验条件下比TV，Sparsity，Softcut取得了更好的结果。进一步分析实验可以发现用结构相似度搜索图像的相似块，使得重构后的图像更加充分地保留了原图像的结构信息。本方法提供了处理图像的新思路，使得对图像质量评价的研究和图像处理的研究可以结合起来共同推动图像处理方法的创新。%Image super -resolution is always an active research area of image processing . This paper proposed an image quality assessment method for image super -resolution .Used the structural similarity measurement for block matching , and made use of the weighted aver-age of similar blocks to reconstruct the new high -resolution image .The experiments showed that our method achieves better super -resolution results than TV , Sparsity and Softcut under the same conditions .The experiment results showed that the method proposed in this paper could remind the structure unchanged .This method provided a new way to handle images . The research in image processing could connect image quality assessment with traditional methods .