改进的稀疏表示图像超分辨率复原算法  

Improved image super-resolution via sparse representation

作　　者：邱大伟[1] 刘彦隆[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《电视技术》2016年第1期135-140,共6页Video Engineering

基　　金：山西省自然科学基金项目(2013011017-3)

摘　　要：针对信号的稀疏分解特征,结合图像的超分辨率复原的特点,提出了基于稀疏表示的图像超分辨率复原算法,对两个过完备字典的训练过程、稀疏表示复原算法处理过程进行阐述,同时对改进算法中采用的优化的特征提取算法和自适应边缘方向插值优化低分辨率图像的初始估计两个过程进行详细描述,并通过MATLAB对其进行仿真和验证,实验结果表明,改进算法的复原效果进一步提高,图像细节能够得到恢复,获得更好的鲁棒性。In view of the feature on signal sparse decomposition, together with the characteristic of the image super-resolution, an algorithm of image super-resolution via sparse representation is designed in this paper. The training process of the two over-complete dictionary and the processing procedure of the resolution via sparse representation is described. And then, the improved algorithm of the optimizing the feature extraction algorithm and optimizing the initial value of low-resolution image with adaptively selects the directional edge direction interpolation are described in detail. Meanwhile, the simulation and verification are also given based on MATLAB. Experimental results show that improved algorithm can effectively extract more superior quality than original algorithm, image can be directly restored. At the same time, the improved algorithm can obtain better robustness.

关 键 词：稀疏表示 特征提取 边缘插值 

分 类 号：TN957.52[电子电信—信号与信息处理]