Multi-stage progressive image super-resolution method

Abstract

The invention discloses a multi-stage progressive image super-resolution method. The method comprises the following steps of extracting the to-be-reconstructed image features by using a multi-scale feature extraction module, introducing a channel attention module, and endowing different channel features with weights so as to enhance the expression capability of the important channel features; adopting a residual feature fusion mechanism, utilizing the relevance of image context features fully, and obtaining a reconstructed image at the first stage; using a refinement module for optimizing the reconstructed image obtained at the first stage so as to obtain a finer reconstructed image; adopting a loss function for training, and further improving the super-resolution effect of the model. According to the method, the effect of image super-resolution reconstruction is effectively improved, and better effects are obtained at the aspects of subjective vision and objective evaluation indexes.

Description

technical field [0001] The invention relates to an image super-resolution method based on deep learning, in particular to a multi-stage progressive image super-resolution method. Background technique [0002] Image super-resolution reconstruction is one of the key technologies in the field of computer vision. Its purpose is to reconstruct a high-resolution (LR) image with more details from one or more existing low-resolution (LR) images. High-Resolution, HR) image. It has a wide range of applications in many fields, such as medical imaging, video surveillance, remote sensing images, etc. Since the higher the spatial resolution, the richer the details of the image, the stronger the ability to express the information, which is beneficial to the subsequent image processing, analysis and understanding. However, in the actual imaging process, due to degradation factors such as atmospheric jitter, optical blur, motion blur, undersampling and noise, there is an inevitable deviati...

AI Technical Summary

Problems solved by technology

[0003] In view of this, the object of the present invention is to provide a multi-stage progressive image super-resolution method, which overcomes the lack of correlation between image context features and the neglect of high-frequency features of the reconstructed image itself in existing image super-resolution methods. problem, thereby improving the reconstruction effect of image super-resolution

Images