P‐8.5: An Efficient Algorithm for Single Image Super‐Resolution

Abstract

In order to solve the problem of insufficient key information and serious damage of detail texture in the feature extraction process of single image super‐resolution reconstruction algorithm when processing low‐resolution image, we propose a high‐performance feature information dependence algorithm (HPDA) based on a Prior Preserving Block (PPB) and Residual Feature Holding Block (RFHB) to obtain feature information dependence between image blocks. To better focus key information in the original image, We introduced an improved Enhanced Spatial Attention module (ESA) to ensure the overall integrity and accuracy of the image reconstruction. The reconstruction quality detail and reconstruction speed of different scales have achieved a good balance.