摘要
为提高现有图像的分辨率,在Yang等提出的基于稀疏表示的图像重建方法的基础上进行改进,提出联合高低分辨率块求取稀疏系数的方法。使用初次重建的高分辨率块和输入的低分辨率块在联合字典下再次稀疏分解,用联合分解出的新系数和高分辨率字典重建最终的高分辨率块,该方法可通过迭代使用提升重建性能。实验结果表明,改进的方法在目视效果和PSNR及SSIM二项指标值上都比Yang等方法更佳。
To improve the resolution of the existing images,basing on the reconstruction method proposed by Yang et al,which is based on sparse represent,an approach combining high and low resolution patches to calculate sparse represent coefficients was presented.The high resolution patches reconstructed initially and the input of low resolution patches were used together to calculate the sparse coefficients in the high-low-resolution-union dictionary,and the final high resolution patches were reconstructed using the high resolution dictionary and the coefficients.This method could be iteratively used for further performance improvement.Experimental results show the proposed method is better than Yang＇s method in both the visual effect and the PSNR and SSIM values.
出处
《计算机工程与设计》
北大核心
2016年第12期3290-3294,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(41161065
40901207)
贵州省科学技术厅
贵州师范大学联合科技基金项目(黔科合J字LKS[2013]28号)
关键词
超分辨率
图像重建
优化方法
稀疏表示
超完备字典
字典训练
super-resolution
image reconstruction
optimization methods
sparse representation
over-complete dictionary
dictionary training
