Lossless compression of hyperspectral images using hybrid context prediction

Area of Science:

• Computer Science
• Image Processing
• Data Compression

Background:

• Lossless compression is crucial for hyperspectral imaging due to large data volumes.
• Existing algorithms face challenges in balancing compression ratio and computational complexity.

Purpose of the Study:

• To propose a novel lossless compression algorithm for hyperspectral images.
• To improve compression efficiency and reduce computational cost compared to existing methods.

Main Methods:

• The algorithm employs a two-stage approach: decorrelation and coding.
• Decorrelation includes intraband (median prediction) and interband (hybrid context prediction) stages.
• Hybrid context prediction combines linear prediction (LP) and context prediction, with residuals entropy coded via arithmetic coding.

Main Results:

• The proposed algorithm achieves high compression ratios.
• It demonstrates low complexity and computational cost.
• Performance is validated against established algorithms like 3D-CALIC, M-CALIC, and JPEG-LS.

Conclusions:

• The developed hybrid context prediction algorithm offers an effective solution for lossless hyperspectral image compression.
• It presents a favorable trade-off between compression performance and computational demands.