Research On Remote Sensing Image Quality Enhancement Method Based On Deep Learning

Remote sensing images,as carriers that can accurately and objectively express the shape of targets and their background information,have been widely used in fields such as meteorological monitoring,urban planning,and military reconnaissance.However,due to limitations in hardware technology and various external factors,the quality of remote sensing images often cannot meet the needs of practical applications,which greatly affects the completion of remote sensing image interpretation tasks.Therefore,how to improve the quality of images in a low-cost and efficient manner has become an area of great scientific research value in the field of remote sensing image interpretation.This article focuses on the key scientific issues in enhancing the quality of remote sensing images,and conducts research on optical remote sensing image denoising and super-resolution reconstruction based on deep learning.The main research contents are as follows:(1)The degradation mechanism of remote sensing images and the influence of degradation factors on image quality are studied.The advantages and disadvantages of traditional methods and deep learning methods in image denoising and super-resolution are compared and analyzed,providing theoretical support for deep learning-based research on enhancing the quality of remote sensing images.(2)A composite noise removal method of “Identification-Denoising” is proposed to address the problem of diverse noise types and difficult removal in remote sensing images.Firstly,a noise category identification network is established to effectively identify various combinations of noise categories.Then,combined with single-class noise removal network and noise reduction method,preliminary denoising results are obtained.Finally,through multi-level wavelet convolutional neural network,residual noise is removed,which improves the denoising performance and generalization ability of the network and achieves effective removal of composite noise in remote sensing images.(3)An area-based super-resolution method is proposed to address the problem of difficulty in super-resolution of regions with different complexity in remote sensing images.Based on identifying the complexity of image areas,this method processes image areas separately using super-resolution sub-networks of different scales,and recombines them into a high-definition image to improve the network’s learning ability for effective information and achieve high-performance remote sensing image super-resolution.The methods proposed in this article have been verified through remote sensing data,and experimental results show that these methods can effectively solve the main problems in enhancing the quality of remote sensing images and provide technical support for enhancing the quality of remote sensing images.