Toggle search
Toggle menu
5.4K
6.1K
230
53.3K
VR & AR Wiki
Toggle preferences menu
Toggle personal menu
Not logged in
Your IP address will be publicly visible if you make any edits.

Anti-aliasing

vrarwiki.com · VR & AR Wiki
More actions

Anti-aliasing (also written antialiasing, and abbreviated AA) is the set of techniques that reduce aliasing artifacts in rendered images: jagged, stair-stepped edges, small or thin objects that break up or disappear, and detail that flickers or crawls as the camera moves. These artifacts come from representing a continuous scene with a finite grid of pixel samples.[1][2] In real-time rendering the term usually means spatial anti-aliasing, and the common methods are supersampling (SSAA), multisample anti-aliasing (MSAA), post-process filters such as FXAA and SMAA, temporal anti-aliasing (TAA), and machine-learning methods such as NVIDIA's DLAA.

In virtual reality (VR), aliasing is more visible than on a monitor because each pixel of a head-mounted display covers a larger angle of the visual field, and because the user's head is always moving.[3][4] Temporal anti-aliasing, the default in many flat-screen games, tends to blur or judder in a headset, so Valve, Oculus (now part of Meta), Epic Games and Unity Technologies have all recommended MSAA with a forward renderer for VR applications (Unity specifically for untethered headsets with tile-based GPUs).[3][5][6][7]

Reviewed 21 October 2026. Checked every claim, quote, number and citation against the cited papers (Crow 1977, PBR book, Yang 2020, Akeley 1993 abstract, Reshetov 2009, FXAA white paper, SMAA, Kaplanyan 2016, Patney 2016, Waldow 2024), the Valve GDC 2015 slides, Oculus, Epic, Meta, Unity, Impulse Gear and NVIDIA pages. About review dates.

Causes of aliasing

Franklin C. Crow's 1977 paper "The aliasing problem in computer-generated shaded images" explained the visible defects of digitally shaded images in terms of aliasing in sampled signals.[1] He wrote that the problems occur in three situations: along silhouette edges and surface creases, in very small objects, and in areas of complicated detail. Because each pixel is a sample of an infinitely small point in the scene, a very small object can fall between samples and vanish, and a long thin object can appear "as a string of beads". In animation the defects become more obvious: Crow described "armies of ants" running along edges as their slopes change, small details flashing on and off, and horizontal edges jumping from one raster line to the next.[1]

The underlying theory is the sampling theorem. In the textbook Physically Based Rendering: From Theory to Implementation, Matt Pharr, Wenzel Jakob and Greg Humphreys state that a signal can be reconstructed exactly only if it is sampled at more than twice its highest frequency (the Nyquist frequency). Geometric edges are step functions with infinite frequency content, so no finite sample rate captures them exactly, and the authors conclude that "we can never remove all of its sources". The textbook separates errors made during sampling (prealiasing) from errors made during reconstruction (postaliasing).[2] Crow named three remedies: raise the resolution, blur or smooth the finished image, or make each sample represent a finite area of the scene instead of a point, which amounts to filtering the scene before it is sampled.[1]

Aliasing is not limited to geometric edges. Shading can alias as well. Kaplanyan and co-authors describe specular aliasing as the undersampling of highly glossy highlights on curved or bumpy surfaces: the highlights can be much smaller than a pixel and carry a lot of energy, so they shimmer when the surface or camera moves.[8] Texture detail, alpha-tested cut-outs (such as foliage) and thin geometry are other common sources, and different anti-aliasing methods handle them differently.[9][10]

History

According to Pharr, Jakob and Humphreys, Crow "first identified aliasing as a major source of artifacts in computer-generated images". They credit Robert Cook, and separately Mark Dippé and Erling Wold, with using nonuniform (stochastic) sampling to turn aliasing into noise in the mid-1980s.[11]

For real-time graphics, a 2020 survey by Lei Yang, Shiqiu Liu and Marco Salvi of NVIDIA cites Kurt Akeley's 1993 SIGGRAPH paper on the RealityEngine graphics system as the reference for hardware-accelerated MSAA, which the authors say was the most used anti-aliasing technique in real-time engines before TAA.[12] Akeley's abstract describes RealityEngine as a system "designed primarily to render texture mapped, antialiased polygons", with a measured fill rate of over 240 million antialiased, texture-mapped pixels per second.[13]

The same survey notes that MSAA is awkward to combine with deferred shading: the hardware mapping between visibility samples and shading is lost, which effectively turns MSAA into costly supersampling, and multisampled G-buffers need much more memory and bandwidth.[12] Because of these difficulties, various post-processing techniques that work on the finished image were proposed to replace MSAA in deferred renderers.[12] Morphological antialiasing (MLAA), which Alexander Reshetov of Intel Labs presented at High Performance Graphics 2009 as a post-processing step for ray tracing; it looks for patterns in a rendered image and blends colors near them using simple rules, without casting extra rays, and works with any rendering technique.[14] Timothy Lottes's FXAA white paper for NVIDIA (version 1.0, January 2011) describes FXAA as "a high performance and high quality screen-space software approximation to anti-aliasing" that runs as a single full-screen pixel shader pass on a single-sample color image.[15] The organizers of the SIGGRAPH 2011 course "Filtering Approaches for Real-Time Anti-Aliasing" wrote that the original CPU-based MLAA method "gave birth to an explosion of real-time anti-aliasing techniques that rival MSAA"; the course brought together twelve talks on such techniques, including MLAA, FXAA 3.11 and implementations for game consoles.[16] SMAA (Jimenez, Echevarria, Sousa and Gutierrez, Eurographics 2012) combined morphological antialiasing with multisampling, supersampling and temporal reprojection.[17]

Temporal anti-aliasing became the dominant method in games during the 2010s. Yang and colleagues date its rise to about 2010, citing Halo: Reach and Crysis 2; NVIDIA offered TXAA from 2012, and engines including Unreal Engine 4 and Unity added TAA as a standard option. By 2020 the survey called TAA "the de facto standard" for anti-aliasing in almost all game engines and real-time 3D renderers.[12] Machine-learning approaches followed. NVIDIA describes its Deep Learning Anti-Aliasing (DLAA) as an AI-based anti-aliasing technique that uses the same Super Resolution technology as DLSS but keeps the native render resolution; NVIDIA lists it as supported on GeForce RTX 20 to 50 series GPUs.[18]

Techniques

Method How it works Main strengths Main weaknesses
Supersampling (SSAA) Renders more samples than output pixels and shades every sample, then filters down Every sample is shaded, so shading detail is anti-aliased along with edges Cost grows with the sample count[9][8]
Multisample anti-aliasing (MSAA) Stores coverage and depth for several samples per pixel but runs the pixel shader once per pixel per primitive, then resolves Extra cost concentrated on edges; efficient on tile-based GPUs Does not fix specular or texture aliasing; awkward with deferred shading[10][12][19]
Morphological and post-process filters (MLAA, FXAA, SMAA) Find edges in the finished image and blend across them Low cost; independent of the rendering technique Cannot restore detail lost between samples; often temporally unstable[14][15][1][12]
Temporal anti-aliasing (TAA) Jitters the sample position each frame and blends reprojected history with the current frame Supersampling at a small cost over single-sample shading; also reduces specular aliasing Soft images, ghosting, temporal lag; blur from head motion in VR[12][6]
Machine-learning reconstruction (for example DLAA) An AI model produces the anti-aliased image at native resolution Uses the same Super Resolution technology as DLSS, without lowering the render resolution Requires a supported NVIDIA GeForce RTX GPU[18]

Supersampling

Supersampling renders the scene at a higher resolution than the target and filters the result down, so each output pixel is computed from several fully shaded samples. Daniel Rákos notes that its performance and memory cost rise with the sample count.[9] In VR, "supersampling" also refers to raising a headset's render resolution above the runtime default; that use is covered in the Supersampling article.

Multisample anti-aliasing

MSAA keeps several depth and coverage samples per pixel but, in its basic form, executes "only a single fragment shader invocation for each pixel a primitive overlaps with". A per-pixel coverage mask records which samples each primitive covers, and the samples are averaged in a resolve step.[9] Because pixels in the interior of a triangle are shaded once, the extra work is concentrated on edges. Meta's developer documentation states that when a triangle covers all sample locations of a pixel, "there are no additional costs compared to rendering without any anti-aliasing".[19]

The same design leaves some aliasing in place. Unity's documentation states that MSAA does not fix shader aliasing "such as specular or texture aliasing".[10] Textures are sampled once per pixel, and the edges of alpha-tested geometry stay aliased; per-sample shading, alpha-to-coverage or separate supersampling are the usual workarounds.[9] A GPU can also be asked to shade every sample, which turns MSAA into supersampling.[9]

Post-process anti-aliasing

Post-process methods analyze the final color image. MLAA infers silhouettes from color discontinuities in the image and blends colors around them.[14][16] FXAA targets aliasing "both on triangle edges and in shader results" and includes logic for sub-pixel aliasing; Lottes gave its cost as under 1 ms for a 1920x1200 frame on a GeForce GTX 480 at preset 2, and noted that it can save memory compared with MSAA, "especially on stereo and multi-display render targets".[15] FXAA is meant to run after tone mapping, on the low-dynamic-range sRGB image.[15] Yang and colleagues note that many post-process methods "suffer from temporal stability problems" because a single frame does not contain enough information to recover the true pixel values.[12]

Temporal anti-aliasing

TAA performs spatial anti-aliasing with data gathered over several frames. A sub-pixel jitter offset shifts the projection each frame; the previous output is reprojected with motion vectors and blended with the new frame, typically with an exponential weight, after a history-validation step that rejects or clamps stale data. With a commonly used blending factor of 0.1, the survey calculates an effective maximum of 19 samples per pixel at steady state.[12] The survey notes that TAA is sometimes criticized for producing soft images, and that history validation errors cause two kinds of artifact: invalid history treated as valid produces ghosting and temporal lag, while valid history treated as invalid produces blur and temporal instability.[12] Unity's documentation likewise warns that TAA "often creates ghosting artifacts in extreme situations".[10] Epic Games writes that "in most cases" TAA is preferable to MSAA "because it removes both geometric aliasing and specular aliasing", although it makes an exception for VR (see below).[6]

Shading and specular anti-aliasing

Because MSAA and post-process filters do little against sub-pixel highlights, renderers also filter the shading itself. One family of methods widens the material roughness where the surface normal varies within a pixel. Valve stored the spread of normals in each normal-map mip level as a roughness value, and added a "geometric roughness" term computed from screen-space derivatives of the interpolated vertex normal, since "dense meshes without normal maps also alias".[3] Kaplanyan, Hill, Patney and Lefohn's 2016 method filters the microfacet normal distribution function over the pixel footprint; they describe it as temporally stable and compatible with deferred shading and normal maps, and compare it with Valve's approach.[8] For Unreal Engine projects that use MSAA, Epic suggests building content to limit specular aliasing, for example with the Normal to Roughness feature.[6]

Anti-aliasing in virtual reality

Why aliasing is more visible in headsets

In his GDC 2015 talk "Advanced VR Rendering", Valve's Alex Vlachos wrote that "Aliasing is your enemy" in VR. He estimated that VR at the time delivered about 15.3 pixels per degree (110 degree field of view with a 1.4x render scale), against about 25 for a 720p 30-inch monitor and about 50 for a 2560x1600 30-inch monitor, both at a 50 degree horizontal field of view. Because "the camera (your head) never stops moving", aliasing is amplified, and Valve's customers described it as "sparkling". One of the goals Valve listed was that aliasing should not be noticeable to customers.[3] The 2024 Waldow study makes the same point: aliasing artifacts are "particularly noticeable" because the pixels of current VR displays subtend a large field of view, and the authors connect jagged edges and temporal flicker with visual discomfort and eye strain.[4]

MSAA and forward rendering

Vlachos treated 4x MSAA as the minimum quality for VR, used 8x MSAA "if perf allows", and stated that "Forward renderers win for antialiasing because MSAA just works". He considered jittered supersampling "obviously the best" when performance could be spared, and advised comparing any image-space method side by side with 4x and 8x MSAA.[3]

In May 2016 Oculus engineer Pete Demoreuille described a forward renderer that Oculus had written for Unreal Engine 4. He wrote that Unreal's TAA "works extremely well in monitor-only rendering" but "causes noticeable artifacts in VR", specifically "judder and geometric aliasing during head motion", and that the team found MSAA "to produce sharper, more detailed images". Forward rendering made MSAA possible and removed full-screen passes that can interfere with asynchronous timewarp. He also noted that screen-space effects can introduce incorrect stereo disparities between the two eyes.[5]

Current engine documentation keeps this guidance. Epic's Forward Shading Renderer page states that in VR "the constant sub-pixel movement introduced by head tracking generates unwanted blurriness" with TAA, which makes MSAA the better choice, while warning that in Epic's tests MSAA instead of TAA increased GPU frame time by about 25 percent and that MSAA projects need content built to limit specular aliasing.[6] Epic's XR best-practices page tells developers to enable Forward Shading and set the anti-aliasing method to MSAA.[20] Meta calls the forward renderer the recommended rendering path for Meta Quest development, says that MSAA can save GPU time compared with TAA, and states that its Meta XR Project Setup Tool recommends 4x MSAA for all Quest devices.[21]

Standalone headsets and tile-based GPUs

Standalone headsets use tile-based mobile GPUs, which changes the cost of MSAA. Meta's documentation explains that the depth buffer is tiled and fits in about 1 MB of on-chip memory; a multisampled buffer splits the frame into more, smaller tiles, "but internal benchmarks show that 4x MSAA is extremely cheap on Quest devices".[19] Meta's MSAA analysis, measured with a 1216 x 1344 eye buffer, put the overhead at 0.5 to 1.5 ms per frame depending on content and GPU level, or about 10 to 15 percent for a medium-complexity app at 4x. It advises developers to "almost always use MSAA" on Quest, not to go above 4x, and to spend remaining GPU time on higher-resolution render targets instead.[22] Because the extra samples never leave tile memory, the multisampled buffers can be lazily allocated in Vulkan; Meta estimates the memory saved at about 90 MB per eye buffer for 4x MSAA at 1680x1760, the Meta Quest 3 default resolution, and about 66 MB per eye buffer at the Meta Quest 2 default of 1440x1584.[23]

Unity gives similar advice for untethered XR devices. Its Unity 6.3 manual says that tile-based GPUs "can store more samples in the same tile", which makes MSAA efficient on mobile and untethered XR platforms, and that 2x MSAA "provides a good balance between visual quality and performance". The same page recommends the forward path, because deferred rendering's G-buffer requires slow graphics-memory loads on tile-based GPUs, and notes that post-processing without on-tile post-processing (available from Unity 6.3) is expensive on these devices because it adds intermediate textures.[7] A 2020 survey of TAA likewise notes that MSAA "is often cheaper than TAA on lower-end platforms (especially tile-based mobile GPUs)".[12]

Exceptions and other approaches

Not every VR title follows the MSAA recommendation. Impulse Gear wrote that while developing Farpoint for PlayStation VR, it moved to the clustered forward renderer Oculus had written for UE4, but found that MSAA "was not only very slow, it was also not very effective at removing the aliasing". The game shipped with FXAA; supersampling "made the game look much better but it was extremely expensive", so the team added dynamic resolution scaling.[24]

Machine-learning anti-aliasing has also reached PC VR. In May 2021 NVIDIA announced the first VR games with DLSS and said that in Into the Radius DLSS brought "a noticeable improvement to anti-aliasing, which greatly reduces shimmering and stair-stepping on objects and foliage".[25]

Anti-aliasing also interacts with foveated rendering, which lowers the sampling rate in the periphery. Anjul Patney and colleagues at NVIDIA reported in 2016 that their foveated renderer filtered both before and after shading "to address aliasing from undersampling in the periphery" and used a "multiresolution- and saccade-aware" temporal antialiasing algorithm; they measured temporal stability similar to temporally filtered non-foveated rendering while reducing the number of shades by up to 70 percent.[26]

Research

Kristoffer Waldow and colleagues compared four anti-aliasing methods built into the Unity engine in a pairwise-comparison user study presented at the 2024 GI VR/AR workshop. The setup used an HP Reverb G2 (native resolution 2160 x 2160 pixels) with the per-eye render resolution raised to 2700 x 2630 to account for lens distortion, an NVIDIA RTX 3090 and Unity 2023.1.0a25. The methods were 2x SSAA, 4x MSAA, TAA (Unity's High setting with contrast-adaptive sharpening of 0.85) and FXAA. Twenty-four participants took part, randomly assigned to two groups of 12, and each scene was shown in a static and a moving variant.[4]

SSAA produced the best image quality and FXAA the worst, and the ranking did not differ significantly between static and moving scenes. SSAA differed significantly from MSAA, but its difference from TAA was not statistically significant, and TAA and MSAA achieved similar results. The authors argued for SSAA in VR while noting its large performance cost at 4x or 8x, and named deep-learning methods such as DLAA as a subject for future work.[4]

See also

References

  1. ↑ 1.0 1.1 1.2 1.3 1.4 Franklin C. Crow (1977-11). "The aliasing problem in computer-generated shaded images". Communications of the ACM, vol. 20, no. 11, pp. 799-805. doi:10.1145/359863.359869. https://doi.org/10.1145/359863.359869. Retrieved 2026-10-21.
  2. ↑ 2.0 2.1 Matt Pharr, Wenzel Jakob, Greg Humphreys. "Sampling Theory". Physically Based Rendering: From Theory to Implementation, 3rd edition. https://pbr-book.org/3ed-2018/Sampling_and_Reconstruction/Sampling_Theory. Retrieved 2026-10-21.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 Alex Vlachos (2015-03). "Advanced VR Rendering". Game Developers Conference 2015. Valve. https://media.steampowered.com/apps/valve/2015/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf. Retrieved 2026-10-21.
  4. ↑ 4.0 4.1 4.2 4.3 Kristoffer Waldow, Jonas Scholz, Martin Misiak, Arnulph Fuhrmann, Daniel Roth, Marc Erich Latoschik (2024). "Anti-aliasing Techniques in Virtual Reality: A User Study with Perceptual Pairwise Comparison Ranking Scheme". GI VR / AR Workshop 2024. Gesellschaft für Informatik e.V.. doi:10.18420/vrar2024_0015. https://dl.gi.de/items/d094d5b9-f6f5-4618-8178-349e257f7d9a. Retrieved 2026-10-21.
  5. ↑ 5.0 5.1 Pete Demoreuille (2016-05-25). "Optimizing the Unreal Engine 4 Renderer for VR". Meta Horizon OS Developers Blog. Meta Platforms. https://developers.meta.com/horizon/blog/introducing-the-oculus-unreal-renderer/. Retrieved 2026-10-21.
  6. ↑ 6.0 6.1 6.2 6.3 6.4 "Forward Shading Renderer in Unreal Engine". Unreal Engine 5.8 Documentation. Epic Games. https://dev.epicgames.com/documentation/en-us/unreal-engine/forward-shading-renderer-in-unreal-engine. Retrieved 2026-10-21.
  7. ↑ 7.0 7.1 "Optimize for untethered XR devices in URP". Unity Manual (Unity 6.3). Unity Technologies. https://docs.unity.com/en-us/engine/6000.3/manual/xr/graphics/untethered-device-optimization.md. Retrieved 2026-10-21.
  8. ↑ 8.0 8.1 8.2 Anton S. Kaplanyan, Stephen Hill, Anjul Patney, Aaron Lefohn (2016). "Filtering Distributions of Normals for Shading Antialiasing". High Performance Graphics 2016. Eurographics Association. doi:10.2312/hpg.20161201. https://research.nvidia.com/sites/default/files/pubs/2016-06_Filtering-Distributions-of/NDFFiltering.pdf. Retrieved 2026-10-21.
  9. ↑ 9.0 9.1 9.2 9.3 9.4 9.5 Daniel Rákos (2021-10-19). "Multisampling primer". RasterGrid. https://www.rastergrid.com/blog/gpu-tech/2021/10/multisampling-primer/. Retrieved 2026-10-21.
  10. ↑ 10.0 10.1 10.2 10.3 "Add anti-aliasing in the Universal Render Pipeline". Unity Manual (Unity 6.2). Unity Technologies. https://docs.unity3d.com/6000.2/Documentation/Manual/urp/anti-aliasing.html. Retrieved 2026-10-21.
  11. ↑ Matt Pharr, Wenzel Jakob, Greg Humphreys. "Sampling and Reconstruction: Further Reading". Physically Based Rendering: From Theory to Implementation, 3rd edition. https://pbr-book.org/3ed-2018/Sampling_and_Reconstruction/Further_Reading. Retrieved 2026-10-21.
  12. ↑ 12.00 12.01 12.02 12.03 12.04 12.05 12.06 12.07 12.08 12.09 12.10 Lei Yang, Shiqiu Liu, Marco Salvi (2020-05). "A Survey of Temporal Antialiasing Techniques". Computer Graphics Forum, vol. 39, no. 2, pp. 607-621 (Eurographics 2020 State of the Art Report). doi:10.1111/cgf.14018. https://doi.org/10.1111/cgf.14018. Retrieved 2026-10-21.
  13. ↑ Kurt Akeley (1993). "RealityEngine graphics". Proceedings of SIGGRAPH '93, pp. 109-116. ACM. doi:10.1145/166117.166131. https://www.microsoft.com/en-us/research/?p=156903. Retrieved 2026-10-21.
  14. ↑ 14.0 14.1 14.2 Alexander Reshetov (2009). "Morphological antialiasing". Proceedings of the Conference on High Performance Graphics 2009, pp. 109-116. ACM. doi:10.1145/1572769.1572787. https://doi.org/10.1145/1572769.1572787. Retrieved 2026-10-21.
  15. ↑ 15.0 15.1 15.2 15.3 Timothy Lottes (2011-01-25). "FXAA". NVIDIA white paper. NVIDIA. http://developer.download.nvidia.com/assets/gamedev/files/sdk/11/FXAA_WhitePaper.pdf. Retrieved 2026-10-21.
  16. ↑ 16.0 16.1 Jorge Jimenez, Diego Gutierrez, et al.. "Filtering Approaches for Real-Time Anti-Aliasing". ACM SIGGRAPH 2011 Courses. https://www.iryoku.com/aacourse/. Retrieved 2026-10-21.
  17. ↑ Jorge Jimenez, Jose I. Echevarria, Tiago Sousa, Diego Gutierrez (2012). "SMAA: Enhanced Subpixel Morphological Antialiasing". Computer Graphics Forum, vol. 31, no. 2, pp. 355-364. doi:10.1111/j.1467-8659.2012.03014.x. https://www.iryoku.com/smaa/. Retrieved 2026-10-21.
  18. ↑ 18.0 18.1 "NVIDIA DLSS". NVIDIA GeForce. NVIDIA. https://www.nvidia.com/en-us/geforce/technologies/dlss/. Retrieved 2026-10-21.
  19. ↑ 19.0 19.1 19.2 "Mobile GPUs and improved algorithms". Meta Horizon OS Developers. Meta Platforms. https://developers.meta.com/horizon/documentation/unity/gpu-improved-algorithms/. Retrieved 2026-10-21.
  20. ↑ "XR Best Practices in Unreal Engine". Unreal Engine 5.8 Documentation. Epic Games. https://dev.epicgames.com/documentation/en-us/unreal-engine/xr-best-practices-in-unreal-engine. Retrieved 2026-10-21.
  21. ↑ "Forward Shading Renderer". Meta Horizon OS Developers. Meta Platforms. https://developers.meta.com/horizon/documentation/unreal/unreal-forward-renderer/. Retrieved 2026-10-21.
  22. ↑ "Multisample Anti-Aliasing Analysis for Meta Quest". Meta Horizon OS Developers. Meta Platforms. https://developers.meta.com/horizon/documentation/native/android/mobile-msaa-analysis/. Retrieved 2026-10-21.
  23. ↑ "OpenXR Meta Quest Support settings". Meta Horizon OS Developers. Meta Platforms. https://developers.meta.com/horizon/documentation/unity/unity-openxr-settings-quest/. Retrieved 2026-10-21.
  24. ↑ Victor Ceitelis, Greg Koreman (2018-03-15). "A Look Inside Farpoint's Rendering Techniques for VR". Impulse Gear Developer Blog. Impulse Gear. https://www.impulsegear.com/a-look-inside-farpoints-rendering-techniques-for-vr. Retrieved 2026-10-21.
  25. ↑ Andrew Burnes (2021-05-18). "NVIDIA DLSS: No Man's Sky And 8 Other Games, Including The First VR Titles, Add Performance-Accelerating Tech This Month". NVIDIA GeForce News. NVIDIA. https://www.nvidia.com/en-gb/geforce/news/may-2021-rtx-dlss-game-update/. Retrieved 2026-10-21.
  26. ↑ Anjul Patney, Marco Salvi, Joohwan Kim, Anton Kaplanyan, Chris Wyman, Nir Benty, David Luebke, Aaron Lefohn (2016-12-05). "Towards Foveated Rendering for Gaze-Tracked Virtual Reality". ACM Transactions on Graphics, vol. 35, no. 6 (SIGGRAPH Asia 2016). Association for Computing Machinery. doi:10.1145/2980179.2980246. https://research.nvidia.com/publication/2016-12_towards-foveated-rendering-gaze-tracked-virtual-reality. Retrieved 2026-10-21.
Retrieved from "https://vrarwiki.com/index.php?title=Anti-aliasing&oldid=52813"