Tobias Alexander Franke

STAR-NT: Spatiotemporal Acceleration of Real-Time Neural Transparency Rendering

Grigoris Tsopouridis et al. and Tobias Alexander Franke Proceedings of Computer Graphics International 2026

Quality comparison of methods: FLIP mean error (white) and MSE * 10<sup>2</sup> (green). The ground truth is shown on the left with the depth complexity (maximum number of transparent layers) in the parenthesis.
Quality comparison of methods: FLIP mean error (white) and MSE * 102 (green). The ground truth is shown on the left with the depth complexity (maximum number of transparent layers) in the parenthesis.

Abstract

Neural order-independent transparency delivers high-quality rendering of overlapping transparent surfaces, but its geometry passes and network input generation remain costly, particularly on mobile and legacy hardware. We present a spatiotemporal acceleration framework that exploits spatial and temporal coherence to reduce this overhead while preserving visual quality. Spatially, we use adaptive quadtree-based screen-space subdivision to scale geometry pass resolution according to local color variance. Temporally, selected frames reuse the previous transparency result through depth-based reprojection instead of full rendering. Together, these optimizations reduce rendering cost and integrate efficiently into existing real-time rendering pipelines.

Preview

Supplemental Video

Links