Hyperspectral image compression: adapting SPIHT and EZW to anisotropic 3-D wavelet coding.

Area of Science:

• Image processing
• Data compression
• Remote sensing

Background:

• Hyperspectral images require efficient compression due to large data volumes.
• Wavelet-based compression offers adaptability and reasonable complexity for such data.
• Existing wavelet methods have been applied to hyperspectral space missions.

Purpose of the Study:

• To optimize a full wavelet compression system specifically for hyperspectral images.
• To define an optimal 3-D wavelet decomposition in a rate-distortion sense.
• To adapt and compare zerotree coding methods (EZW, SPIHT) and JPEG 2000.

Main Methods:

• Developed an algorithm for optimal 3-D wavelet decomposition.
• Identified a fixed decomposition with comparable performance and lower complexity.
• Adapted Extended Zerotree Wavelet (EZW) and Set Partitioning In Hierarchical Trees (SPIHT) algorithms.
• Compared performance against JPEG 2000 on diverse hyperspectral datasets.

Main Results:

• A specific fixed wavelet decomposition was found to be near-optimal and significantly better than isotropic decomposition.
• This decomposition enables efficient zerotree coding algorithms.
• Adapted EZW and SPIHT methods showed competitive performance compared to JPEG 2000.

Conclusions:

• The optimized wavelet compression system offers improved efficiency for hyperspectral images.
• The proposed fixed decomposition and zerotree coding provide a practical and effective compression solution.
• This work contributes to better data management for hyperspectral remote sensing applications.