[Compression of interference hyperspectral image based on FHALS-NTD].

Area of Science:

• Signal Processing
• Data Compression
• Hyperspectral Imaging

Context:

• Hyperspectral interference images contain rich spectral information but are data-intensive.
• Existing compression methods struggle with the high dimensionality and complexity of hyperspectral data.
• Efficient compression is crucial for storage, transmission, and analysis of hyperspectral datasets.

Purpose:

• To propose a novel hyperspectral interference image compression algorithm.
• To leverage Fast Hierarchical Alternating Least Squares Nonnegative Tensor Tucker Decomposition (FHALS-NTD) for improved compression.
• To evaluate the algorithm's reliability, stability, and compressive properties.

Summary:

• The proposed algorithm decomposes hyperspectral interference images using 3-D OPD lifting-based discrete wavelet transform (3D OPT-LDWT).
• The decomposed sub-bands are treated as a nonnegative tensor and processed by the FHALS-NTD algorithm.
• Quantization and bit-plane coding are applied to core tensors and component matrices to generate the compressed bit-stream.

Impact:

• The algorithm demonstrates reliable and stable performance with good compressive properties.
• Achieves an average peak signal-to-noise ratio (PSNR) above 40 dB across compression ratios from 32:1 to 4:1.
• Outperforms traditional methods by improving average PSNR by 1.23 dB, enhancing hyperspectral image compression efficiency.