基于L0范数和稀疏编码的单幅图像超分辨率重建方法

Single-image super-resolution reconstruction methodbased on L0 norm and sparse coding

基于稀疏编码的方法在单幅图像的超分辨率重建中获得了一定的成功,但是这类方法却存在着重建出错误的边缘和重建的图像块之间差异性的丢失等问题。为了解决这些问题,认为一幅高分辨率图像是由边缘成分和纹理成分两部分组成,提出了一种基于L0范数和非局部拉普拉斯稀疏编码的单幅图像超分辨率重建方法。首先,为了能够重建出正确的图像边缘,提出了一种基于L0范数的针对图像边缘的超分辨率重建方法;然后,在纹理成分的超分辨率重建阶段,提出了一种非局部的拉普拉斯稀疏编码(NLSC)来实现图像纹理成分的超分辨率重建;最后,试验结果表明,提出的方法能够有效解决现有方法中存在的问题,获得更高质量的高分辨率图像。

Methods based on sparse coding have been successfully used in single-image super-resolution reconstruction. However, there are problems about incorrect edges and difference loss among the reconstructed image patches. To overcome these problems, we assume that a high resolution image consists of two components: the edge component and the texture component, hence propose a new approach based on L0 norm and non-local Laplacian sparse coding for single-image super-resolution reconstruction. First, in order to correctly reconstruct image edges, we propose a super-resolution reconstruction method based on L0-norm image edges. Second, for the texture component super-resolution reconstruction, a non-local Laplacian sparse coding is proposed. Finally, experiment results demonstrate that the proposed approach can overcome incorrect edges and difference loss effectively, and achieve more competitive single-image super-resolution quality.