Image super-resolution reconstruction method based on multi-scale pyramid network

Abstract

The invention discloses an image super-resolution reconstruction method based on a multi-scale pyramid network. The image super-resolution reconstruction method comprises the following steps of S1, performing shallow feature extraction on an input image; S2, performing feature fusion and feature enhancement on the shallow features through K multi-scale residual modules to obtain richer deep features; S3, performing up-sampling on the deep features by using transposed convolution; S4, reconstructing the image by using residual learning; and S5, taking the reconstructed image as the output of the current pyramid network and the input of the next layer of pyramid network, and continuing to adopt the steps S1-S4 for training, so as to obtain an image with higher resolution. According to the method, a multi-scale residual module is adopted to fuse features to obtain richer features; meanwhile, a Laplacian pyramid network is adopted to gradually upsample and reconstruct a high-resolution image; and by means of the method, images with richer details and higher quality can be reconstructed.

Description

technical field [0001] The invention relates to the technical fields of computer vision and image processing, in particular to an image super-resolution reconstruction method based on a multi-scale pyramid network. Background technique [0002] With the development of information technology, the number of pictures on the Internet is constantly increasing. As a main medium for people to perceive the world, images have been applied to various scenarios. In many fields, as large as the field of medical images, satellite remote sensing, as small as people's cameras, mobile phones, etc. People have higher and higher requirements for image quality. Therefore, improving the resolution of images will be of great significance in real life. [0003] Image super-resolution reconstruction aims to restore high-resolution images from one or more low-resolution images, and has become one of the research hotspots in the field of computer vision in recent years. At present, super-resoluti...

AI Technical Summary

Problems solved by technology

However, the current mainstream methods are based on the theory that the deeper the network, the better the reconstruction effect. With the increase of the network depth, problems such as gradient disappearance or grid degradation still exist, and most of the methods are up-sampled once to achieve The specified size, so the quality of the reconstructed high-resolution image needs to be improved

Images