A Survey of Temporal Antialiasing Techniques

Abstract

Temporal Antialiasing (TAA), formally defined as temporally‐amortized supersampling, is the most widely used antialiasing technique in today's real‐time renderers and game engines. This survey provides a systematic overview of this technique. We first review the history of TAA, its development path and related work. We then identify the two main sub‐components of TAA, sample accumulation and history validation, and discuss algorithmic and implementation options. As temporal upsampling is becoming increasingly relevant to today's game engines, we propose an extension of our TAA formulation to cover a variety of temporal upsampling techniques. Despite the popularity of TAA, there are still significant unresolved technical challenges that affect image quality in many scenarios. We provide an in‐depth analysis of these challenges, and review existing techniques for improvements. Finally, we summarize popular algorithms and topics that are closely related to TAA. We believe the rapid advances in those areas may either benefit from or feedback into TAA research and development.

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

“…Note that the blurring occurs because the reprojection generally results in fractional pixel locations, so resampling is required. The resampling is typically done via bilinear or bicubic filtering [18]. Nevertheless, a reasonable amount of blur is still advantageous, as it acts as a simple denoising filter and usually improves the quality of the result, as we see in Section IV.…”

“…The accumulated samples are reprojected onto the new frame, and then blended with new samples according to some blending factor α. The common approach of using a constant α results in an exponential smoothing filter, also known as exponential moving average, assigning progressively lower weights to the older samples that get less relevant [18].…”

“…With TAA, the camera is typically sub-pixel jittered with a Halton sequence for achieving the same smoothing effect even with a static camera. For more details on TAA, we refer to the recent survey [18]. However, note that TAA can introduce an excessive amount of blur.…”

A reduced-precision network for image reconstruction

“…Moreover, if the rejection heuristic is too aggressive the temporal accumulation becomes ineffective. Yang et al [2020] survey recent TAA techniques and provide an in-depth analysis of the image quality trade-offs with these heuristics.…”

“…TAA uses renderer-generated motion vectors to gather and accumulate samples from previous frames, effectively increasing the number of samples per pixel. However, TAA is susceptible to ghosting artifacts, loss of detail, and temporal instability [Yang et al 2020].…”