Variable-length input super-resolution video reconstruction method based on deep learning

Abstract

The invention discloses a variable-length input super-resolution video reconstruction method based on deep learning. The method comprises the following steps: constructing a training sample with a random length, and obtaining a training set; establishing a super-resolution video reconstruction network model, wherein the super-resolution video reconstruction network model comprises a feature extractor, a gradual alignment fusion module, a depth residual error module and a superposition module which are connected in sequence; training the super-resolution video reconstruction network model by adopting the training set to obtain a trained super-resolution video reconstruction network; and sequentially inputting to-be-processed videos into the trained super-resolution video reconstruction network for video reconstruction to obtain corresponding super-resolution reconstructed videos. According to the method, a gradual alignment fusion mechanism is adopted, alignment and fusion can be carried out frame by frame, and alignment operation only acts on two adjacent frames of images, so that the model can process a longer time sequence relationship, more adjacent video frames are used, that is to say, more scene information is contained in input, and the reconstruction effect can be effectively improved.

Description

Technical field [0001] The invention belongs to the technical field of video restoration, and in particular relates to a method for reconstructing super-resolution video with variable length input based on deep learning. Background technique [0002] Most applications based on images and videos depend on the quality of the image. In general, the quality of an image is related to the amount of information it contains. The image resolution is used to measure the amount of information contained in an image. It is expressed by the number of pixels per unit area, such as 1024× 768. It can be seen that the resolution of the image represents the quality of the image, so in real life and application scenarios, high resolution becomes the quality requirement of images and videos. [0003] However, when the video contains complex motions such as occlusion, severe blur, and large offsets, it is necessary to reconstruct the video to obtain high-quality video information. In order to effecti...

AI Technical Summary

Problems solved by technology

The use of variable-length input sequences solves the problem of inaccurate alignment of long input image sequences in video super-resolution tasks; the use of gradual alignment fusion networks can align and fuse any number of adjacent frames without affecting subsequent reconstruction tasks, making it more practical powerful

Images