Pyramid Attention "Zero-Shot" Network for Single-Image Superresolution.

Xianjun Han,Huabin Wang,Xuejun Li,Hongyu Yang

DOI: https://doi.org/10.1109/tnse.2022.3192471

IF: 6.6

2022-01-01

IEEE Transactions on Network Science and Engineering

Abstract:Single-image superresolution (SISR) is one of the requisite image processing methods used to reconstructs a high-resolution (HR) image from a low-resolution (LR) observation. Existing SISR methods mostly rely on supervised learning and are confined to specific training data, where implicit conformity matching of the LR images with their high-resolution (HR) counterparts is conducted. However, real LR images rarely obey these constraints, and the degradation process is much more intricate and unknown. In this paper, we proposed a pyramid attention zero-shot (PAZS) network for SISR that sufficiently explores the information hidden in an image itself by learning the patch distribution at a different scale of the image. The proposed pyramid generative model learns the patch distribution in the internal patch learning stage. Then, based on the learned intrinsic property, we explore the corresponding superresolution (SR) image by integrating the intrinsic information into a self-attention mechanism with a progressive generation style at test time. This mechanism consists of internal-external attention and a cross-scale guided fusion module as the connecting passageway. Thus, the PAZS architecture maintains both the global structure and the fine textures of the SR image. This allows SR to be performed without other training datasets and can be adapted to different settings for each image. Experiments on diverse datasets demonstrate that the proposed method outperforms other methods based on single-image training.