Transform coding for hardware-accelerated volume rendering.

Area of Science:

• Computer Graphics
• Image Processing
• Scientific Visualization

Background:

• Hardware-accelerated volume rendering on GPUs is standard for real-time applications.
• Limited graphics memory restricts rendering of large volume datasets.
• Existing volumetric compression methods are often too slow for real-time hardware rendering.

Purpose of the Study:

• To develop a novel volumetric compression technique optimized for real-time hardware-accelerated volume rendering.
• To address the challenge of limited graphics memory when rendering large datasets.
• To achieve fast decompression without sacrificing compression quality.

Main Methods:

• A novel block-based transform coding scheme is proposed.
• Decompression is coupled with rendering, consolidating inverse transform and dequantization.
• Reprojection is precomputed, and encoding is optimized offline.
• A new block classification scheme preserves perceptually important features.

Main Results:

• The developed scheme enables very large volumes to be compressed.
• Decompression is achieved in real-time during GPU rendering.
• The technique maintains compression quality while ensuring fast decompression.
• Perceptually important features are preserved through block classification.

Conclusions:

• The proposed asymmetric transform coding scheme effectively enables real-time rendering of large, compressed volume datasets on GPUs.
• This method overcomes memory limitations in hardware-accelerated volume rendering.
• The approach balances compression efficiency with the demands of real-time visualization.