Accelerating In-transit Isosurface Generation With Topology Preserving Compression

Yanliang Li,Jieyang Chen

2024-08-10

Abstract:Data visualization through isosurface generation is critical in various scientific fields, including computational fluid dynamics, medical imaging, and geophysics. However, the high cost of data sharing between simulation sources and visualization resources poses a significant challenge. This paper introduces a novel framework that leverages lossy compression to accelerate in-transit isosurface generation. Our approach involves a Compressed Hierarchical Representation (CHR) and topology-preserving compression to ensure the fidelity of the isosurface generation. Experimental evaluations demonstrate that our framework can achieve up to 4x speedup in visualization workflows, making it a promising solution for real-time scientific data analysis.

Graphics,Distributed, Parallel, and Cluster Computing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in scientific data visualization, especially during the isosurface generation process, how to accelerate in - transit data processing through the use of topology - preserving compression techniques. Specifically, the paper aims to address the following challenges:

1. **High - cost data sharing**: The cost of data sharing between scientific simulations and visualization resources is very high, especially when performing I/O operations through the file system.

2. **Bottlenecks in data volume and transmission speed**: With the progress of scientific instruments and simulation codes, the amount of data generated and the speed increase dramatically, causing network - based data streams to become a bottleneck in in - transit visualization.

3. **Requirement for real - time data analysis**: In order to achieve real - time or near - real - time scientific data analysis, an efficient visualization workflow is required to provide timely user feedback.

To solve these problems, the paper proposes a new framework that utilizes lossy compression techniques to accelerate isosurface generation in transit. Specific methods include:

- **Compressed Hierarchical Representation (CHR)**: Pre - process and compress data into the CHR format for easy and rapid isosurface generation.
- **Topology - preserving compression**: Ensure that the decompressed compressed data can still generate the same isosurface as the original data, thereby guaranteeing the fidelity of isosurface generation.

Through these methods, the paper shows that its framework can achieve a speed - up of up to 4 times in the visualization workflow, making it a promising solution for real - time scientific data analysis.

### Formula Explanation

In the topology - preserving compression mentioned in the paper, calculating the error bound to ensure topological invariance is a crucial step. The specific formulas are as follows:

Suppose \( k \) is the isovalue of the isosurface, and \( s_0 \) and \( s_1 \) are the values of two vertices, and \( s_0 < k < s_1 \), then the errors \( d_0 \) and \( d_1 \) are respectively:

\[ d_0=|k - s_0| \]
\[ d_1=|k - s_1| \]

To ensure topological invariance, the error must satisfy:

\[ \text{error}<\min(d_0, d_1) \]

Therefore, the minimum error bound \( \epsilon \) can be calculated through the following steps:

1. Find all edges that satisfy \( s_0 < k < s_1 \).
2. Calculate \( d_0 \) and \( d_1 \) for each edge and save the results in an array \( D \).
3. The minimum value in the array \( D \) is the absolute error bound \( \epsilon \).

For the case where a certain proportion of topological errors are allowed (for example, tolerating 5% of cell errors), the \( n \) - th smallest value in the array \( D \) can be selected as the error bound.