基于稀疏表示与线性回归的图像快速超分辨率重建

发布时间：2018-04-03 18:07

本文选题：线性回归　切入点：超分辨率　出处：《智能系统学报》2017年01期

【摘要】：单幅图像超分辨率的目的是从一幅低分辨率的图像来重构出高分辨率的图像。基于稀疏表示和邻域嵌入的超分辨率图像重建方法使得重建图像质量有了极大的改善。但这些方法还很难应用到实际中,因为其重建图像的速度太慢或者需要调节复杂的参数。目前大多数的方法在图像重建的速度和质量两个方面很难有一个好的权衡。鉴于以上问题提出了一种基于线性回归的快速图像超分辨率重建算法,将稀疏表示和回归的方法有效地结合在一起。通过稀疏表示训练的字典,用一种新的方式将整个数据集划分为多个子空间,然后在每一类子空间中独立地学习高低分辨率图像之间的映射关系,最后通过选择相应的投影矩阵来重建出高分辨图像。实验结果表明,相比于其他方法,本文提出的算法无论在图像重建速度还是重建质量方面都取得了更好的超分辨率重建效果。
[Abstract]:The purpose of super-resolution of a single image is to reconstruct a high-resolution image from a low-resolution image.The super-resolution image reconstruction method based on sparse representation and neighborhood embedding greatly improves the image quality.However, these methods are difficult to be applied in practice because the reconstruction speed is too slow or complex parameters need to be adjusted.At present, most methods are difficult to balance the speed and quality of image reconstruction.In view of the above problems, a fast super-resolution image reconstruction algorithm based on linear regression is proposed, which combines sparse representation with regression effectively.Through sparse representation training dictionary, the whole data set is divided into multiple subspaces in a new way, and then the mapping relationship between high and low resolution images is studied independently in each kind of subspace.Finally, the high resolution image is reconstructed by selecting the corresponding projection matrix.The experimental results show that compared with other methods, the proposed algorithm achieves better super-resolution reconstruction results in terms of image reconstruction speed and reconstruction quality.
【作者单位】：北京交通大学信息科学研究所;北京交通大学现代信息科学与网络技术北京市重点实验室;
【基金】：国家自然科学基金项目(61272028,61572067) 国家“863”计划项目(2014AA015202) 广东省自然科学基金项目(2016A030313708) 北京市自然科学基金项目(4162050)
【分类号】：TP391.41

【相似文献】