摘要
为了获取高质量的超分辨遥感图像,提出了一种改进特征提取算子的稀疏表示遥感图像超分辨率重建方法。该算法通过设置模板,对一阶和二阶梯度滤波算子进行改进,在有效提取低分辨率图像边缘特征的同时,减少噪声干扰。利用遥感图像库训练得到高、低分辨率图像块字典,再应用低分辨率图像块及其字典求出稀疏表示系数。高、低分辨率图像块具有相同的稀疏表示系数,可根据已求的稀疏表示系数得出超分辨重建图像。实验结果表明,改进算法的超分辨重建效果的客观评价指标,比以往稀疏表示超分辨方法有很大提高,峰值信噪比提高近0.24dB,均方根误差降低近0.15。
In order to obtain high quality remote sensing images,a sparse representation remote sensing image super-resolution reconstruction algorithm is proposed based on an improved feature extraction operator.The algorithm can improve the first order and second order gradient filter operators by setting template parameters.It can not only extract the edge features of low resolution image effectively but also reduce the noise.High-low resolution image block dictionaries are obtained by training the remote sensing image database.Then the sparse representation coefficient is solved by the image block with low resolution and its dictionary.The high-low resolution image blocks have the same sparse representation coefficients,so the super-resolution reconstruction image can be obtained according to the sparse representation coefficient.Experiment results show that the objective evaluation parameters of the improved SRR algorithm are higher than those of the previous methods.The peak signal-to-noise ratio is increased about 0.24 dB,and the root mean square error is decreased about 0.15.
作者
黄鑫
朱福珍
王成全
巫红
HUANG Xin;ZHU Fuzhen;WANG Chengquan;WU Hong(College of Electrical Engineering,Heilongjiang University,Harbin 150080,China;School of Computer Science and Electrical Engineering,East University of Heilongjiang,Harbin 150080,China)
出处
《黑龙江大学自然科学学报》
CAS
2019年第5期618-623,共6页
Journal of Natural Science of Heilongjiang University
基金
国家自然科学基金资助项目(61601174)
黑龙江省博士后科研启动基金项目(LBH-Q17150)
黑龙江省普通高等学校电子工程重点实验室(黑龙江大学)开放课题资助及省高校科技创新团队资助项目(2012TD007)
黑龙江省省属高等学校基本科研业务费基础研究项目(KJCXZD201703)
黑龙江省自然科学基金资助项目(F2018026)
黑龙江省普通本科高等学校青年创新人才培养计划项目(UNPYSCT-2018012)
关键词
稀疏表示
图像超分辨率
字典训练
特征提取
sparse representation
image super-resolution
dictionary training
feature extraction
