An image super-resolution reconstruction method based on deep learning

Abstract

The invention discloses an image super-resolution reconstruction method based on deep learning, the purpose of which is to use deep learning technology to train low-resolution data to obtain a mapping function between low-resolution and high-resolution, and the technical key lies in ( 1) Downsampling the data set; (2) Using the residual principle, adding the convolution activation results between different layers; (3) The training data is divided into two types: labeled and unlabeled, and the two cases correspond to two (4) Integrate the two types of situations to obtain the final loss function. The invention inputs any low-resolution image into the trained neural network model, and the output of the neural network is the reconstructed super-resolution image. The invention effectively improves the quality of the acquired image without changing the hardware equipment of the imaging system.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to an image super-resolution reconstruction method based on deep learning. Background technique [0002] Image super resolution (super resolution, SR) is the process of obtaining a high-resolution image from a low-resolution image. This technology is mainly used to enhance the spatial resolution of the image, which can break through the limitations of the original system imaging hardware conditions. Re-acquired high-resolution images have the characteristics of higher resolution, more detailed information, and higher quality picture quality, and are currently one of the most effective and lowest-cost ways to obtain high-precision images. [0003] In the process of image acquisition, limited by factors such as imaging conditions and imaging methods, the imaging system usually cannot obtain all the information in the original scene, and will be affected by many factors such as deformat...

AI Technical Summary

Problems solved by technology

[0006] (1) SR is an inverse problem. For a low-resolution image, there may be many different high-resolution images corresponding to it

[0007] (2) The low-quality images obtained in real multimedia applications are often complex degraded images with multiple degraded factors, such as low resolution, out-of-focus blur, motion blur, compression distortion and noise, etc.

Images