4

Just finished a little proof of concept of a reprojected multisample antialiasing technique and daaaaamn it's looking good. First time ever a rendering technique I invented isn't complete shit so that's an improvement

It still has some (pretty big) issues with both spatial and temporal ghosting but I have some ideas in the pipeline

I wish I could show you guys comparison images but as it turns out most anti aliasing techniques look pretty good on still frames but only the good ones stand up under motion and I don't know of a good way right now to capture pixel perfect clips like that

Comments
  • 2
    Yup. I normally don’t care that much about anti aliasing but I hate it when the slightest movement of the camera causes this visual noise that instantly catches your attention. So I prefer AA techniques that reduce this noise even if they make the image slightly blurry.
    And of course, one important property is also how computationally expensive it is. Unless you need it to render videos.
  • 2
    @Lensflare Yeah performance is a tricky one. Mine is based on multisampling because even though it is more expensive I strongly believe you cannot get away with only one sample per pixel. With the geometric complexity of todays games you just *need* more geometric information per pixel than one measily sample

    But I don't believe in deferred rendering anyhow so it'll be alright 🤷‍♂️
  • 1
    @Lensflare you don't need AA if your screen is high-res enough ;)
  • 1
    @tosensei I wish this was true but sadly, the noise doesn’t go away even with very high res.
    You might not see individual pixels but you see the effect of pixels changing along the edge of a polygon rapidly.
  • 1
    A higher resolution is essentially what is being simulated with SSAA (super sampling anti aliasing)
Add Comment