Blockwise Multi-Order Feature Regression for Real-Time Path-Tracing Reconstruction

Abstract

Path tracing produces realistic results including global illumination using a unified simple rendering pipeline. Reducing the amount of noise to imperceptible levels without post-processing requires thousands of samples per pixel (spp), while currently it is only possible to render extremely noisy 1 spp frames in real time with desktop GPUs. However, post-processing can utilize feature buffers, which contain noise-free auxiliary data available in the rendering pipeline. Previously, regression-based noise filtering methods have only been used in offline rendering due to their high computational cost. In this article we propose a novel regression-based reconstruction pipeline, called Blockwise Multi-Order Feature Regression (BMFR), tailored for path-traced 1 spp inputs that runs in real time. The high speed is achieved with a fast implementation of augmented QR factorization and by using stochastic regularization to address rank-deficient feature data. The proposed algorithm is 1.8× faster than the previous state-of-the-art real-time path-tracing reconstruction method while producing better quality frame sequences.

A Survey of Temporal Antialiasing Techniques

“…It then enables optimal convergence rate at the cost of an additional storage channel. This method is used in some applications that require fast convergence after disocclusion [WMB19, KIM∗ 19], particularly when target frame rate is low.…”

“…Reprojection‐based temporal data reuse has also been widely applied in denoising ray‐traced images. Many techniques rely on TAA‐like temporal filter to denoise path‐traced images rendered at very low sampling rates [SKW∗17, MMBJ17, SPD18, KIM∗ 19, LLCK19, WMB19, SA19, BBHW∗19]. One major problem of reusing samples in path‐tracing is the difficulty to obtain correct motion vectors for secondary shading effects like reflections and lighting.…”

Systematic Evaluation of the Quality Benefits of Spatiotemporal Sample Reprojection in Real-Time Stereoscopic Path Tracing

“…Sample reprojection is particularly relevant for real-time path tracing because of the strict time budget available for rendering each frame. Applying reprojection and temporal accumulation is roughly an order of magnitude faster than real-time path tracing [13], [31], so it is a cheap way to increase the quality of the rendered result without heavy computational overhead.…”

“…Nowadays reprojection is used by modern real-time path tracing reconstruction filters [13]- [16], in order to generate a better quality input before the actual denoising or reconstruction filter. Moreover, reprojection is used extensively in rasterized game graphics.…”

Neural Temporal Adaptive Sampling and Denoising

“…The quality is plausible, if somewhat lower than the learned approaches, but performance is attractive for interactive applications: a 1920 × 1080 frame is denoised in a few milliseconds. A similar approach using blockwise regression was recently proposed by Koskela et al [KIM∗19].…”