A fast super-resolution image reconstruction algorithm based on NLDCT feature fusion

摘要

The traditional neighbor embedding-based super-resolution image reconstruction algorithm has several disadvantages such as visible blocking artifacts and high computational complexity. This paper combines the image patch Normalization Luminance (NL) feature and Discrete Cosine Transform (DCT) transformation feature for preserving image detail information and de-blocking artifacts. And the K-means method is used to cluster the training image set for reducing the computational complexity of neighbor selection in the process of neighborhood reconstruction. So a Fast Super-resolution image Reconstruction algorithm based on NL&DCT feature fusion (FSR-NL&DCT) is proposed. Experimental results on MIT image database show that the proposed algorithm improves the quality of reconstructed super-resolution images, which the PSNR average value is increased 1.0551dB and the SSIM average value is also increased 0.0613.