Technology Category
Machine Learning: Learning on the Edge & Model CompressionCognitive Modeling & Cognitive Systems: Neural Spike CodingComputer Vision: Low Level & Physics-based Vision
Collaborative Texture Filtering
🤖 AI Summary
Modern texture compression formats cannot leverage GPU hardware texture units for decompression and filtering, leading to visual artifacts—including noise, flickering, and appearance distortion—in stochastic texture filtering (STF). To address this, we propose a wave-level cooperative texture filtering method that exploits GPU wave intrinsics to share decoded texel values among neighboring threads, eliminating redundant decompression. By integrating cooperative thread execution, spatial reuse, and an enhanced spatiotemporal denoising scheme, our approach achieves zero-error filtering at ≤1 texel sample per pixel and provides high-fidelity fallback under low-magnification scenarios. Experiments demonstrate substantial reduction in decompression overhead, complete elimination of STF’s inherent noise and flicker, and visual quality approaching lossless rendering. This work establishes a new paradigm for high-quality, real-time filtering of highly compressed textures in graphics applications.
Technology Category
Application Category
📝 Abstract
Recent advances in texture compression provide major improvements in compression ratios, but cannot use the GPU's texture units for decompression and filtering. This has led to the development of stochastic texture filtering (STF) techniques to avoid the high cost of multiple texel evaluations with such formats. Unfortunately, those methods can give undesirable visual appearance changes under magnification and may contain visible noise and flicker despite the use of spatiotemporal denoisers. Recent work substantially improves the quality of magnification filtering with STF by sharing decoded texel values between nearby pixels (Wronski 2025). Using GPU wave communication intrinsics, this sharing can be performed inside actively executing shaders without memory traffic overhead. We take this idea further and present novel algorithms that use wave communication between lanes to avoid repeated texel decompression prior to filtering. By distributing unique work across lanes, we can achieve zero-error filtering using <=1 texel evaluations per pixel given a sufficiently large magnification factor. For the remaining cases, we propose novel filtering fallback methods that also achieve higher quality than prior approaches.
Problem
Research questions and friction points this paper is trying to address.
Improve texture filtering quality with GPU wave communication
Reduce texel decompression overhead in magnification scenarios
Minimize visual noise and flicker in stochastic methods
Innovation
Methods, ideas, or system contributions that make the work stand out.
GPU wave communication for texel sharing
Zero-error filtering with unique lane work
Novel fallback methods for higher quality
🔎 Similar Papers
No similar papers found.
