🎉 Celebrating 25 Years of GameDev.net! 🎉

Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!

Hybrid motion blur

Started by
3 comments, last by sensate 11 years, 6 months ago
Lookdev test for future Hind

[media]http://vimeo.com/54064796[/media]

I have been lucky enough to be working with Brian Burrell's Mi-27r model over the past month.
It has been incredibly useful in finding some holes regarding common game asset development patterns.
brianburrell.com/3dper_mi27r.html

This video (and still image) is a look development, effects and animation proof of concept pass for a small realtime effects demo that I'm planning to complete early in the new year.

It demonstrates a new type of hybrid motion blur using geometry shader fin extraction and image space motion blur paired together.
While there are still undesirable artifacts to fix I wanted to make it public to show a "proof of concept" of the idea that I will be writing about further in the coming months.

The motion blur is a hybrid of geometry shader fin extrusion and the recently published reconstruction filter for plausible motion blur)
http://graphics.cs.w...otionBlurI3D12/

Click here to view the iotd
Advertisement
Neat, looks good so far! I'm sure there are some artefacts, but I didn't see any in a single viewing.
Thanks :). The artifacts are in the geometry shader extrusion stage.
This is caused by bad transform interpolation through a looping animation.
To fix this I have to rethink some parts of my scene graph and animation system.
Interesting. I'm very curious how you're combining the two techniques, and what tradeoffs you're making. I'm guessing that you're trying to increase quality via the geometric technique, while still retaining some of the performance from our image-space algorithm?
Interesting. I'm very curious how you're combining the two techniques, and what tradeoffs you're making. I'm guessing that you're trying to increase quality via the geometric technique, while still retaining some of the performance from our image-space algorithm?

Yes. The geometry technique fills in areas that would otherwise be unshaded, while your algorithm gives me the cues for blurring.

The geometry stage uses an OIT technique similar to the linked list algorithm presented by AMD a couple of years ago.

This topic is closed to new replies.

Advertisement