Video super-resolution method based on adversarial learning and attention mechanism

Abstract

The invention discloses an end-to-end video super-resolution method based on adversarial learning and an attention mechanism in order to overcome defects that a traditional video resolution method ishigh in calculation overhead, low in calculation efficiency and incapable of efficiently processing a long sequence. According to the method, space-time correlation is extracted by adopting adjacent frame fusion and an attention mechanism, and a long sequence is processed at a time by adopting a circulating structure, so that a high-resolution reconstructed video rich in details and coherent in time sequence can be obtained. The video super-resolution method based on the attention mechanism and the adversarial learning has the advantages: 1, the novel video super-resolution method based on theattention mechanism and the adversarial learning is provided, and the super-resolution effect is improved; 2, the video super-resolution method based on the attention mechanism and the adversarial learning provided by the invention is better in effect; and 3, the video super-resolution can be applied to an actual scene, such as a monitoring device and a satellite image.

Description

technical field [0001] The invention belongs to the technical field of computer digital image processing, and in particular relates to a video super-resolution method based on an attention model and an adversarial learning model. Background technique [0002] The main way for humans to obtain information is vision, and most vision-based applications depend on image quality. However, usually due to factors such as hardware equipment or harsh environments, it is difficult to obtain high-resolution video images. Super-resolution technology processes a given sequence of low-resolution images or video frames to reconstruct detailed high-resolution images or video frames without the cost of upgrading the imaging system. [0003] The early super-resolution technology was proposed in the 1980s. Initially, mathematical methods were used for reconstruction, such as iterative and repeated projection methods and interpolation methods, but failed to achieve good results. At present, th...

AI Technical Summary

Problems solved by technology

Second, the sliding window used in motion estimation contains a large amount of computational redundancy. Only a single frame can be reconstructed by using multiple frames each time. The same frame will participate in the calculation several times, and the calculation efficiency for long sequence videos is low.

Images