DLSS, FSR, TSR
https://zhuanlan.zhihu.com/p/598916763
https://github.com/GPUOpen-Effects/FidelityFX-FSR2
output quantization error
UE TAA output 最后加了一个很小的noise
Stocastically round up or down using the hardware RWTexture2D truncation unit to take into precision loss due to pixel format encoding.
1 | taa_half3 QuantizeForFloatRenderTarget(taa_half3 Color, taa_half E, float3 QuantizationError) |