Coupled Dense Convolutional Neural Networks with Autoencoder for Unsupervised Hyperspectral Super-Resolution
Abstract
This paper proposes CoDenNet (Coupled Dense Convolutional Network), an unsupervised deep learning approach for hyperspectral image super-resolution reconstruction. By coupling dense convolutional layers with an autoencoder architecture, our method achieves superior performance in reconstructing high-resolution hyperspectral images from low-resolution inputs.
Key Contributions
- Developed an unsupervised learning framework for hyperspectral super-resolution
- Integrated dense convolutional networks with autoencoder for feature extraction
- Achieved state-of-the-art performance on benchmark datasets
- Presented at ICIG 2023 conference
Keywords
Hyperspectral imaging, Super-resolution, Deep learning, Convolutional neural networks, Unsupervised learning
摘要
本文提出 CoDenNet，一种面向高光谱图像超分辨率重建的无监督深度学习方法，通过耦合密集卷积网络与自编码器，从低分辨率输入中重建高分辨率高光谱图像。
方法要点
- 构建无监督高光谱超分辨率重建框架。
- 结合密集卷积网络与自编码器提取特征。
- 相关工作发表于 ICIG 2023，完整实验与结果以论文原文为准。
关键词
高光谱成像、超分辨率、深度学习、卷积神经网络、无监督学习。