Toward Feature-Preserving Vector Field Compression.

Area of Science:

• Scientific Visualization
• Data Compression
• Computational Science

Background:

• Topological features in vector fields are essential for scientific understanding.
• Existing lossy compression methods often alter or lose critical point information.
• Accurate preservation of critical points is vital for reliable data analysis.

Purpose of the Study:

• To develop error-bounded lossy compression methods for preserving topological features in 2D and 3D vector fields.
• To specifically focus on the accurate preservation of critical points (location and type).
• To ensure no false positives, negatives, or type errors in decompressed data.

Main Methods:

• Adapting a vertex-wise error bound for each grid point.
• Compressing input data alongside the error bound field using a modified lossy compressor.
• Developing methods for piecewise linear and bilinear vector fields.

Main Results:

• The proposed method successfully preserves critical points within defined error bounds.
• Achieved high compression ratios while maintaining topological accuracy.
• Demonstrated embarrassingly parallelizable compression for large datasets and in situ processing.

Conclusions:

• The developed error-bounded lossy compression effectively preserves critical points in vector fields.
• This method offers a significant improvement over existing lossy compressors for topological feature preservation.
• The technique is suitable for large-scale scientific applications requiring high fidelity.