본문 바로가기
  • Home

Efficient Volume Rendering Using Precomputed Density Queries and Predictive Break Conditions

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(10), pp.81~90
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : August 29, 2025
  • Accepted : September 29, 2025
  • Published : October 31, 2025

Jong-Hyun Kim 1

1인하대학교

Accredited

ABSTRACT

Volume rendering of large 3D density fields remains computationally expensive, limiting real-time use in medical imaging, meteorology, and scientific visualization. We present an efficient method that integrates Precomputed Density Queries with Predictive Break Conditions. During ray casting, the precomputed queries skip unnecessary voxels, while the predictive break conditions estimate when to terminate accumulation by forecasting frame-to-frame density changes via gradient-vector analysis with multi-frame weighting. Implemented in PyCUDA, the approach executes gradient estimation and ray operations on the GPU, minimizing CPU–GPU transfers. Experiments on dynamic gaseous volumes demonstrate up to 20× speedup over conventional baselines while keeping the maximum density deviation ≤ 0.03, yielding artifact-free, temporally coherent results suitable for interactive visualization. The framework is readily applicable to CT/MRI stacks, large-scale weather fields, and fluid/smoke simulations that require high-throughput, real-time rendering.

Citation status

* References for papers published after 2024 are currently being built.

This paper was written with support from the National Research Foundation of Korea.