摘要： 为解决单幅图像的超分辨重建问题，提出一种基于聚类的单帧图像超分辨率重建方法。从高分辨率样本图像中学习一个结构聚类型的高分辨率字典，利用迭代收缩算法优化目标方程，求得高分辨率图像的表示系数，使用学习到的高分辨率字典对低分辨率图像进行重构。实验结果表明，与总变分方法、软切割方法和稀疏表示方法相比，该方法的单帧图像超分辨率重建效果较好。

关键词: 超分辨率, 稀疏表示, 重建方法, 聚类, 字典, 迭代

Abstract: To the question of single frame image super-resolution, implement single frame image super-resolution with the prior of training images, this paper proposes a single frame image super-resolution reconstruction method based on clustering. It builds a structural clustering based high-resolution dictionary from a set of high-resolution images, optimizes objective equation by using iterative shrinkage solution to solve the representation coefficient of high-resolution image, reconstructs low-resolution image by exploiting the learned high-resolution dictionary. Experimental results show that compared with Total Variation(TV) method, Softcuts method and Sparse Representation(SR) method, the effect of the single frame image super-resolution reconstruction of this method is better.

Key words: super-resolution, sparse representation, reconstruction method, clustering, dictionary, iteration

中图分类号: 

• TN911.73