人工智能培训

搜索

人工智能教程:利用空间谱深度残余卷积神经网络进行高光谱图像去噪(Hyperspectral Image Denoising Employing a Spatia

[复制链接]
1yuqunchenggong 发表于 2018-6-4 08:35:14 | 显示全部楼层 |阅读模式
1yuqunchenggong 2018-6-4 08:35:14 977 0 显示全部楼层
人工智能教程:利用空间谱深度残余卷积神经网络进行高光谱图像去噪(Hyperspectral Image Denoising Employing a Spatial-Spectral Deep Residual  Convolutional Neural Network)高光谱图像(HSI)去噪是提高后续HSI解释和应用性能的关键预处理过程。本文提出了一种新的基于深度学习的方法,通过学习非线性端到端映射嘈杂和干净的HSI之间采用组合空间谱深度卷积神经网络(HSID-CNN)。空间和光谱信息都被同时分配给建议的网络。另外,分别采用多尺度特征提取和多尺度特征表示来捕获多尺度空间光谱特征,并将特征表征与不同水平融合用于最终修复。另外,为了保持学习过程的稳定性和效率,重构输出用残余模式表示,而不是直接结果。模拟和实际数据实验表明,所提出的HSID-CNN在定量评估指标,视觉效果和HSI分类准确性方面均实现了许多主流方法。
Hyperspectral image (HSI) denoising is a crucial preprocessing procedure toimprove the performance of the subsequent HSI interpretation and applications.In this paper, a novel deep learning-based method for this task is proposed, bylearning a non-linear end-to-end mappingbetween the noisy and clean HSIs witha combined spatial-spectral deep convolutional neural network (HSID-CNN).Boththe spatial and spectral information are simultaneously assigned to theproposed network.In addition, multi-scale feature extraction and multi-levelfeature representation are respectively employed to capture both themulti-scale spatial-spectral feature and fuse the feature representations withdifferent levels for the final restoration.In addition, to maintain thestability and efficiency of the learning procedure, the reconstructed output isrepresented with residual mode instead of straightforward results.Thesimulated and real-data experiments demonstrate that the proposed HSID-CNNoutperforms many of the mainstream methods in both the quantitative evaluationindexes, visual effects, and HSI classification accuracy.人工智能教程:利用空间谱深度残余卷积神经网络进行高光谱图像去噪(Hyperspectral Image Denoising Employing a Spatial-Spectral Deep Residual  Convolutional Neural Network) VtTBMMgHht0qMlV5.jpg
URL地址:https://arxiv.org/abs/1806.00183     ----pdf下载地址:http://arxiv.org/pdf/1806.00183    ----人工智能教程:利用空间谱深度残余卷积神经网络进行高光谱图像去噪(Hyperspectral Image Denoising Employing a Spatial-Spectral Deep Residual  Convolutional Neural Network)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

1yuqunchenggong当前离线
新手上路

查看:977 | 回复:0

快速回复 返回顶部 返回列表