应用深度卷积神经网络的细节模糊图像重建  

Detail blurred image reconstruction using deep convolutional neural network

作　　者：刘青 俞王杨 孙昊宇 LIU Qing;YU Wangyang;SUN Haoyu(Hubei Central China Technology Development of Electric Power Co.,Ltd.,Wuhan 430000,China)

机构地区：[1]湖北华中电力科技开发有限责任公司,湖北武汉430000

出　　处：《电子设计工程》2025年第6期183-186,191,共5页Electronic Design Engineering

基　　金：湖北华中电力科技开发有限责任公司科研项目(0127DY112073)。

摘　　要：当前细节模糊图像重建方法存在图像清晰度低、图像重建效果差等问题,因此提出应用深度卷积神经网络的细节模糊图像重建方法。利用损失函数对细节模糊图像分辨率进行优化,并根据优化结果调整图像参数。基于图像参数调节,建立细节图像处理的深度卷积神经网络,计算图像重构参数与神经网络状态之间的关系,得到关键特征信息匹配参数计算结果,根据计算结果实现细节模糊图像重建。实验结果表明,应用所提方法重建的图像清晰度在95%以上,重建后细节更加丰富,图像重建效果更好。The current method for reconstructing detail blurred images has problems of low image clarity and poor image reconstruction results.Therefore,a method for reconstructing detail blurred images using deep convolutional neural networks is proposed.Optimize detail blurred images using a loss function and adjust image parameters based on the optimization results.On the basis of adjusting image parameters,establish a deep convolutional neural network for detail image processing,calculate the relationship between image reconstruction parameters and neural network state,obtain the calculation results of key feature information matching parameters,and achieve detail blurred image reconstruction based on the calculation results.The experimental results show that the image clarity reconstructed by the proposed method is above 95%,and the reconstructed details are richer,resulting in better image reconstruction results.

关 键 词：深度卷积神经网络 细节模糊图像 图像重建 图像参数调节 匹配参数 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]