Research On Infrared Image Denoising And High Resolution Reconstruction

Compared with visible light,infrared light has the advantages of strong anti-interference,good penetration and all-weather.Widely used in public security,military,medicine,agriculture and other fields.Infrared image is the image formed by the thermal radiation of the object itself,independent of external light source.The first infrared spectrometers appeared in the early 19 th century.At present,infrared imaging system has been developed to the third generation.However,due to the process limitation of infrared imaging hardware equipment,infrared image has some problems such as low resolution and noise pollution,and the imaging visual effect needs to be improved.Therefore,infrared image denoising and enhancement is worth further study.At present,there are two ways to improve the imaging quality of infrared images: one is to improve the performance of thermal infrared equipment,but the high performance of equipment means high cost;The other is to use software to improve the imaging quality of infrared image,which is both economical and effective.In order to improve the image quality of infrared image,the image denoising algorithm and image super-resolution reconstruction algorithm are studied deeply in this paper.The main work is as follows:(1)The research status of image denoising technology and super-resolution reconstruction technology at home and abroad is introduced.Based on the principle of infrared imaging,the imaging characteristics of infrared image and the main factors affecting the resolution of infrared image are analyzed,and the degradation model of infrared image is discussed.(2)In view of the problem of high noise pollution and low contrast of infrared image,the idea of side window filtering and pre-filtering is applied to infrared image denoising.The basic principle of side window filtering is analyzed.Aiming at the problem that the non-local average denoising effect is not ideal in the case of mixed noise,a non-local average denoising algorithm based on median edge window prefiltering is proposed.The advantages of the improved algorithm are analyzed theoretically,and the denoising performance of each algorithm is compared by simulation.The results show that the infrared image processed by the denoising algorithm in this paper has the best visual effect,and the peak signal-to-noise ratio and structural similarity ratio of the image after denoising are improved by 23.93 and 0.8945,respectively.(3)In order to enhance the resolution of infrared images,a convolutional neural network algorithm based on multi-level jumper depth residual is proposed.The idea of residual network and jumper connection is used to improve the reuse rate of feature information,enhance signal propagation,and solve the problems of gradient disappearance and gradient explosion in deep convolutional networks.Channel and spatial attention mechanisms are used to improve the representation of the model,and conc AT module is used to fuse local feature information.Simulation results show that compared with the traditional CNN hyperclassification network,the improved convolutional neural network reconstruction evaluation index is significantly improved.