Synthetic hyperspectral array video database with applications to cross-spectral reconstruction and hyperspectral...

Area of Science:

• Computer Vision
• Signal Processing
• Data Science

Background:

• Hyperspectral video data is crucial for advanced applications but difficult to acquire with ground truth.
• Existing algorithms often do not fully exploit temporal correlations in hyperspectral video.
• A standardized database is needed for robust algorithm evaluation.

Purpose of the Study:

• Introduce a synthetic hyperspectral video database for algorithm evaluation.
• Propose and evaluate novel algorithms for hyperspectral video processing.
• Demonstrate the database's utility in diverse applications.

Main Methods:

• Generation of a synthetic hyperspectral video database with associated depth maps.
• Extension of a cross-spectral image reconstruction algorithm to incorporate temporal correlation.
• Development of a hyperspectral video coder leveraging temporal information.

Main Results:

• The enhanced cross-spectral reconstruction algorithm achieved up to 5.6 dB increase in peak signal-to-noise ratio (PSNR).
• The novel hyperspectral video coder demonstrated rate savings of up to 10%.
• Results validate the database's effectiveness for evaluating hyperspectral video algorithms.

Conclusions:

• The synthetic hyperspectral video database facilitates rigorous algorithm assessment.
• Exploiting temporal correlations significantly enhances hyperspectral video reconstruction and compression.
• The database supports diverse applications in hyperspectral imaging and video processing.