Task decoupling-based parametric image super-resolution method and system

Abstract

The invention provides a parametric image super-resolution method based on task decoupling, and the method comprises the steps: generating the super-resolution of an input image, carrying out the feature extraction of the input image through a deep convolution network, and generating a high-resolution image lacking detail textures; extracting texture of the reference image, extracting features of the input reference image through a deep convolutional network, performing feature alignment with the high-resolution image, and extracting detail texture information in the reference image; and migrating the texture to the output image, calculating the similarity between the input image and the reference image, migrating detail texture information to the high-resolution image according to the similarity, and generating the high-resolution image with high-frequency texture details. According to the method, parametric super-resolution is decoupled into two tasks, namely a super-resolution task for the input image and a texture migration task for the reference image, the input image and the reference image are processed respectively, and the content-related reference image can be utilized more sufficiently while the negative influence of the irrelevant reference image is eliminated.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a parametric image super-resolution method and system based on task decoupling. Background technique [0002] The goal of image super-resolution is to reconstruct the input low-resolution image into a corresponding high-resolution image. Image super-resolution tasks are widely used in image enhancement, video surveillance, and remote sensing imaging. Therefore, image super-resolution It has received great attention both in academia and industry. [0003] In recent years, the rapid development of deep learning has brought the performance of image super-resolution to a new height. However, due to the ill-posed nature of the super-resolution task itself, the image super-resolution method based on deep learning at this stage will be over-smooth when zooming in at high magnifications. The problem is that although perceptual loss and confrontation loss can improve the...

AI Technical Summary

Problems solved by technology

Although the parametric super-resolution method has a significant improvement in performance compared with the traditional image super-resolution method, the current method still has the problem of insufficient or even misuse of the reference image, which will lead to the loss of the output image. become blurred or even have visual blemishes

Images