基于改进的特征提取方法和稀疏表示的单幅图像超分辨率重建算法

Single Image Super-Resolution Reconstruction Algorithm Based on Improved Feature Extraction and Sparse Representation

为了改善单幅图像的超分辨率重建效果,在基于过完备字典图像的超分辨率重构算法的架构上,应用改进的高斯Laplace算子来提取低频图像的特征,应用于图像重建。该算子主要用于在重建图像的预处理阶段,有效地提取各个方向的边缘特征,既不会造成漏检,也不会加重噪声。实验表明,与现有的几类算法相比较,使用该算子提取特征后,重建图像的效果无论在峰值信噪比还是结相似性都有所提高。

In order to improve the super-resolution reconstruction effect of single image, uses an improved Gaussian Laplace operator to extract the features of low-frequency on image for the image super-resolution reconstruction algorithm based on over-complete dictionary. This oper-ator is mainly used to extract the edge features of each direction effectively during the preprocessing phase of the reconstructed image,which will not result in missed detection and no increase of noise. Experiments show that, compared with the existing algorithms, the ef-fect of reconstructing the image using the operator is improved both in the PSNR and the SSIM.