Deep learning-based super-resolution image reconstruction method and system

Abstract

The invention discloses a deep learning-based super-resolution image reconstruction method and system. The method comprises the steps of acquiring an image to be reconstructed and training data; inputting the training data into a multilayer convolutional neural network based on a residual structure for learning; reconstructing an optimal model acquired through input learning of the image to be reconstructed to acquire a super-resolution image. By performing deep learning through multilayer convolution based on the residual structure, the acquired optimal model can be high in super-resolution image reconstruction ability; by structuring the optimal model through a deep learning method, excessive smoothing of images acquired through an interpolation method can be avoided, and meanwhile, high-resolution images restored through the optimal model can be clear and sufficient in high-frequency details; frequency domain methods can be saved, and lack of correlation of frequency domain data canbe avoided. The deep learning-based super-resolution image reconstruction system comprises a target acquisition unit, a training unit and an image reconstruction unit and can achieve advantageous effects identical to those of the deep learning-based super-resolution image reconstruction method.

Description

technical field [0001] The present invention relates to the field of super-resolution image reconstruction, in particular to a super-resolution image reconstruction method and system based on deep learning. Background technique [0002] With the development of digital technology, camera technology has also made great progress, and people's requirements for picture pixels are getting higher and higher. The traditional approach is to choose a high-pixel camera to take high-resolution photos. However, this does not solve the problem of converting low-resolution pictures into high-resolution pictures. Therefore, the method of image super-resolution came into being. [0003] Image super-resolution methods in the prior art include: a reconstruction method based on interpolation, a reconstruction method based on learning, and a frequency domain reconstruction method. [0004] Interpolation methods such as the nearest element method, bilinear interpolation method and cubic interpol...

AI Technical Summary

Problems solved by technology

[0004] Interpolation methods such as the nearest element method, bilinear interpolation method and cubic interpolation method, etc., the above method for the value of each pixel on the image is calculated and approximated by several points around it, the disadvantage is that the obtained image is too large Smooth, lost a lot of high frequency detail

Learning-based method: A large number of high-resolution image construction learning libraries are used to generate a learning model, and the prior knowledge obtained by the learning model is introduced in the process of restoring low-resolution images to obtain high-frequency details of the image; the disadvantages are Use only the surface features of the image

Frequency domain method: It is an important method in image super-resolution reconstruction, the most important of which is the anti-aliasing reconstruction method; the anti-aliasing reconstruction method is to improve the spatial resolution of the image through anti-aliasing to achieve super-resolution restoration ; The disadvantage is the lack of correlation of frequency domain data

Images