基于聚类字典的图像超分辨率重构算法

针对图像超分辨率(SR)重构在空间邻域选取过程中 细节特征易被大幅度特征分量淹没的问题,提出 一种基于聚类字典的SR重构(DD-NE)算法。图像SR重构是利用信号处理方 法来提高图像分辨 率,针对NE算法在空间邻域选取时细节信号易被大幅度信号淹没的问题,对输入图像及邻域 利用聚类字典进行 稀疏分解。从大、小幅值表示系数中分别重构大、小幅度特征子图,保护邻域计算中的小幅 度特征,并将 低分辨率(LR)图像库及输入图像使用聚类字典表示。细节信号以字典原子的形式得到表达 ,空间邻域度 量转换为字典原子间的度量,从而细节特征对邻域的选择更加准确。实验结果表明,相对于 NE算法,本文算法图像SR 重构的峰值信噪比(PSNR)值平均提升了1.1dB,有效改善了重构效果；重构时间仅为NE算法的30.9%。

Image super-resolution reconstruction algorithm based on directional clustering dictionary

Because detail features may be coverd by the feature components with large ampli tude in process of choosing spatialneighborhood in image super-resolution (SR) reconstruction,a direction ditictionary and neighbor embedding based (DD-NE) SR reconsturction algorithm is proposed. image SR reconstruction using signal processing method The image resolution is improved in.The input image and the neighborhood are sparsely decomposed by using the clustering dictionary.The feature subgraphs with large and small amplitudes are reconstructed from the lar ge and small amplitude values,respectively,The low resolution (LR) image librar y and the input image are represented by a clustering dictionary.The detail signal is expressed in the fo rm of dictionary atoms,and the metric of spatial neighborhood is converted into the measure of dictionary atoms .Experimental results show that compared with NE algorithm,DD-NE reconstruction of algorithm can improve the av erage PSNR by 1.1dB,which greatly improves the reconstruction performance,but reconstruction time is only 30.9% of that of NE algorithm.