Temporal Anti-Aliasing (TAA)
How to use more samples?
- Distributing / reuse samples across frames (time)
- Jittered sampling
![https://s3-us-west-2.amazonaws.com/secure.notion-static.com/422c8cf8-3e43-41f9-9bcc-195a53206963/Untitled.png](https://s3-us-west-2.amazonaws.com/secure.notion-static.com/422c8cf8-3e43-41f9-9bcc-195a53206963/Untitled.png)
如果场景是静止的,就好像是做了4个采样
随机的效果并不见得好,这里是用一个pattern,分布到四个角落。
![https://s3-us-west-2.amazonaws.com/secure.notion-static.com/3b75d9ee-ed71-448f-a63d-0c86b6691b3f/Untitled.png](https://s3-us-west-2.amazonaws.com/secure.notion-static.com/3b75d9ee-ed71-448f-a63d-0c86b6691b3f/Untitled.png)
State of the art image based anti-aliasing solution
- SMAA (Enhanced subpixel morphological AA)
- History: FXAA → MLAA (Morphological AA) → SMAA
![https://s3-us-west-2.amazonaws.com/secure.notion-static.com/27e7e945-bcef-47aa-bda4-f4cc4caf139a/Untitled.png](https://s3-us-west-2.amazonaws.com/secure.notion-static.com/27e7e945-bcef-47aa-bda4-f4cc4caf139a/Untitled.png)
Note: G-buffers should never be anti-aliased!
Temporal Super Resolution
- Super resolution (or super sampling)
- Literal understanding: increasing resolution
- Source 1 (DLSS 1.0): out of nowhere / completely guessed
- Source 2 (DLSS 2.0): from temporal information
- Key idea of Deep Learning Super Sampling (DLSS) 2.0
- Yet another TAA-like application
- Temporally reuse samples to increase resolution
![https://s3-us-west-2.amazonaws.com/secure.notion-static.com/7acc030d-83bf-4665-8139-e38947040ad0/Untitled.png](https://s3-us-west-2.amazonaws.com/secure.notion-static.com/7acc030d-83bf-4665-8139-e38947040ad0/Untitled.png)
Deferred Shading
- Originally invented to save shading time