Research And Application Of Super-resolution Reconstruction Technology Based On Infrared Image

Infrared imaging technology is widely used,but its low image resolution limits its development,so it is necessary to study the super-resolution reconstruction of infrared images.In order to solve the above problems,this thesis proposes a single-frame infrared image super-resolution reconstruction algorithm ISRGAN.In addition,algorithms based on single-frame infrared images cannot utilize information between adjacent frames in the video.Therefore,based on the ISRGAN algorithm,this thesis proposes a multi-frame infrared image super-resolution reconstruction algorithm,which adds the information supplement of adjacent frames to the target frame to obtain the information correlation between multiple frames.The main research contents of this article are summarized as follows:(1)Aiming at the current problem of SRGAN algorithm in infrared image super-resolution reconstruction,a new algorithm ISRGAN is designed,the core is to improve the network structure based on SRGAN algorithm,respectively to improve the model generation network,confrontation network and loss function.The generating network combining the SRGAN algorithm's generating network and the bicubic interpolation algorithm to maintain the information of the low-frequency region of the image;using the relative discriminating network to replace the original standard discriminating network of the SRGAN algorithm to make the generated image high-frequency texture details more detailed;The perceptual loss function of the SRGAN algorithm adds a mean square error loss to achieve a balance between a clear visual effect and higher PSNR and SSIM values.The improvement of each module is analyzed and verified by experiments.Finally,the overall algorithm is subject to experiment and analysis on subjective and objective evaluation indicators.(2)For the problem that the super-resolution reconstruction of single-frame infrared images cannot make full use of the different high-frequency detail information between adjacent multiple frames in the video,combined with the single-frame infrared image super-resolution reconstruction algorithm ISRGAN designed in Chapter 3,A super-resolution reconstruction algorithm based on generating multi-frame infrared images of the adversarial network is designed.The algorithm includes an optical flow estimation algorithm module and a multi-frame infrared image fusion and reconstruction module.The multi-frame infrared image fusion and reconstruction module is based on Chapter 3 to verify the effective ISRGAN algorithm on a single frame infrared image for improved design.Analyze the design of each module separately,and finally subjectively and objectively experiment and analyze the effect of the overall algorithm.(3)Design and implement a super-resolution reconstruction system that can integrate the algorithms of Chapters 3 and 4 and display the images before and after reconstruction in the display window.This software provides a good human-computer interaction experience,enabling users to more easily and quickly use this algorithm to reconstruct low-resolution single-frame images or multi-frame images to obtain high-resolution images.High-resolution images are displayed and compared.