ARM-Net: A Tri-Phase Integrated Network for Hyperspectral Image Compression.

Area of Science:

• Remote Sensing
• Computer Vision
• Data Compression

Background:

• Current hyperspectral image compression methods face high computational complexity due to the large number of spectral bands.
• Existing techniques struggle with performance under resource-constrained environments.

Purpose of the Study:

• To develop an efficient and high-fidelity hyperspectral image compression framework.
• To address the computational challenges of existing methods while preserving spectral information.

Main Methods:

• A triple-phase hybrid framework (ARM-Net) is proposed.
• Adaptive band selection is used to reduce computational load.
• High-fidelity compression of sampled band clusters and a reconstruction network for loss compensation are employed.

Main Results:

• ARM-Net demonstrates significant improvements over state-of-the-art methods on seven hyperspectral datasets.
• Achieved 1-2 dB higher peak signal-to-noise ratio (PSNR) and multiscale structural similarity index measure (MS-SSIM).
• Reduced the average spectral angle mapper (SAM) by approximately 0.1.

Conclusions:

• The proposed ARM-Net framework effectively balances compression efficiency and reconstruction quality for hyperspectral images.
• ARM-Net offers a viable solution for hyperspectral image compression in resource-limited scenarios.