Single-Image Super-Resolution Algorithm Based on Structural Self-Similarity and Deformation Block Features

To solve the problem of insufficient sample resources and poor noise immunity in single-image super-resolution (SR) restoration procedure, the paper has proposed the single-image SR algorithm based on structural self-similarity and deformation block features (SSDBF). First, the proposed method constructs a scale model, expands the search space as much as possible, and overcomes the shortcomings caused by the lack of a single-image SR training sample; Second, the limited internal dictionary size is increased by the geometric deformation of the sample block; Finally, in order to improve the anti-noise performance of the reconstructed picture, a group sparse learning dictionary is used to reconstruct the pending image. The experimental results show that, compared with state-of-the-art algorithms such as bicubic interpolation (BI), sparse coding (SC), deep recursive convolutional network (DRCN), multi-scale deep SR network (MDSR), super-resolution convolutional neural network (SRCNN) and second-order directional total generalized variation (DTGV). The SR images with more subjective visual effects and higher objective evaluation can be obtained through the proposed method. Compared with existing algorithms, the structural network converges more rapidly, the image edge and texture reconstruction effects are obviously improved, and the image quality evaluation, such as peak signal-noise ratio (PSNR), root mean square error (RMSE), and structural similarity (SSIM), are also superior and popular in image evaluation.