基于针对性字典的压缩图像稀疏超分辨率重建

Sparse representation-based super-resolution reconstruction based on targeted dictionary for compressed image

为了有效地重建压缩低分辨率图像,提出一种基于针对性字典的压缩图像稀疏超分辨率重建算法。首先,根据压缩低分辨率图像的形成特点,对训练库图像进行针对性的下采样压缩编码处理,进行超完备字典的训练;然后,通过训练所得的针对性字典对压缩低分辨率图像进行稀疏表示的超分辨率重建。为进一步恢复图像的高频信息,进行了针对性残差字典训练,并对图像进行高频信息补偿,得到稀疏重建后的图像主观效果更加突出,客观评价参数也得到较大提升。实验结果表明,该算法对压缩图像的超分辨率重建更具针对性,具有良好鲁棒性和高效性。

In order to improve the super-resolution reconstruction performance of compressed low-resolution image, this paper proposes a sparse representation-based reconstruction algorithm by using targeted dictionary. Firstly,the dictionary training is aimed at the characteristics and forming of the compressed image, which training library is acquired the targeted process to make a more effective over-complete dictionary for the super-resolution. Then, the compressed low-resolution image is reconstructed by the targeted dictionary spares representation-based super-resolution. Meanwhile, for getting more detail information, the targeted residual dictionary is designed to compensate the loss of high-frequency information. Supported by the targeted dictionary, the method produces excellent SR results on a variety of images, and it achieves a significant improvement in PSNR, SSIM and competitive performance in visual quality. Extensive experiments manifest the robustness and efficiency of the proposed sparse representation-based super-resolution reconstruction based on targeted dictionary for compressed image.