Single-image Super-resolution Reconstruction Method, Apparatus, Device, and Readable Storage Medium

By constructing an end-to-end image super-resolution reconstruction model based on ordinary convolution and intraconvolution, the problem of insufficient flexibility caused by large amounts of parameters in the prior art is solved, efficient image super-resolution reconstruction is achieved, and image quality is improved.

CN113850721BActive Publication Date: 2025-07-29SHAANXI NORMAL UNIV
1 Cites 0 Cited by

Patent Information

Application Number
CN202111129980.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-09-26
Publication Date
2025-07-29
Estimated Expiration
2041-09-26

AI Technical Summary

Technical Problem

The existing single-image super-resolution reconstruction algorithm based on deep learning has a large number of parameters, which leads to insufficient flexibility, making it difficult to efficiently perform super-resolution reconstruction in practical applications.

Method used

The end-to-end image super-resolution reconstruction model is adopted based on the combination of ordinary convolution and intraconvolution. By constructing the image block training data set, the residual feature distillation module and the intraconvolution mechanism are used to reduce the amount of model parameters, and train through stochastic gradient descent and backpropagation algorithms to optimize the model hyperparameters to improve reconstruction performance.

Benefits of technology

While reducing the amount of model parameters, it significantly improves the performance of image super-resolution reconstruction, improves the preservation effect of image details and textures, and is suitable for practical applications.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a single-image super-resolution reconstruction method, apparatus, device, and readable storage medium. A training data set is obtained, and the images in the training data set are cropped to obtain an image patch training data set. An end-to-end image super-resolution reconstruction model based on the combination of ordinary convolution and inner convolution is constructed according to the image patch training data set. Hyperparameters of the end-to-end image super-resolution reconstruction model are set, and after setting a loss function, an optimizer, and a training termination condition, the end-to-end image super-resolution reconstruction model is trained to obtain a trained end-to-end image super-resolution reconstruction model. A low-resolution image is input into the trained end-to-end image super-resolution reconstruction model to obtain a high-resolution image output by the trained end-to-end image super-resolution reconstruction model. The purpose of the present invention is to improve the performance of image super-resolution reconstruction while reducing the number of network parameters.
Need to check novelty before this filing date? Find Prior Art