摘要
基于稀疏表示的超分辨率重构算法效果依赖于样本图像信息,难以保证重构质量;基于图像结构自相似的算法利用了图像自身的附加信息,但是这些信息不足以获得很好的重构效果.本文综合利用样本图像信息和待处理低分辨率图像自身信息,提出了一种新的方法.在基于稀疏表示的框架下把与待重建图像相似的高分辨率样本图像信息提取出来用于重构,利用低分辨率图像自身的附加信息对上一步的重构图像进行修复,进一步提高重构质量.数值实验结果表明,本算法对图像的细节部分具有更好的重构效果.
The reconstructed image quality of Super-resolution based on sparse representation depends on the information of high-resolution image database, the result can not be guaranteed. Super-resolution based on image structural similarity only uses＇ the additional information contained in the given low-reso- lution image itself, but the information is not enough to get an ideal reconstructed image. In order to use both training database and the given low-resolution image, a new method is put forward in this paper： First, use the sparse representation based algorithm to reconstruct the image; and then, use the addition- al information contained in the low resolution image hance the quality in advance. The simulation result two algorithm mentioned above. to repair the achieved image in the first step, en- indicates that the method perform better than the
出处
《纺织高校基础科学学报》
CAS
2013年第4期548-552,共5页
Basic Sciences Journal of Textile Universities
关键词
超分辨率重构
稀疏表示
附加信息
自相似学习
super-resolution
sparse representation
additional information
self-similarity learning
