Mitchell Netravali

[zeitmeister]

Hi guys,
altough I'm using a super sampling of 5(!) I've got the impression that the "anti-aliasing" is not fine enough.
Edges appear to be rough and too sharp; it's hard to say on what my impression depends.

Now I want to beg you for a little more information about the splat- and downsize-filter settings; especially using "Mitchell Netravali Cubic".
What values can I change here to get a finer result in supersampling / anti-aliasing? Even if it might not make sense a all?
I read the technical reference, but I'm not getting too smart out of it. How can "ring", "blurriness" and "radius" help me to get finer results in rendering?

I hope you understand my question; my impression is the same either using PT-BiDir or MLT.


Thank you very much in advance!

[StompinTom]

Hi guys,
altough I'm using a super sampling of 5(!) I've got the impression that the "anti-aliasing" is not fine enough.
Edges appear to be rough and too sharp; it's hard to say on what my impression depends.

Now I want to beg you for a little more information about the splat- and downsize-filter settings; especially using "Mitchell Netravali Cubic".
What values can I change here to get a finer result in supersampling / anti-aliasing? Even if it might not make sense a all?
I read the technical reference, but I'm not getting too smart out of it. How can "ring", "blurriness" and "radius" help me to get finer results in rendering?

I hope you understand my question; my impression is the same either using PT-BiDir or MLT.


Thank you very much in advance!

I find using 'Radial' for Splat and 'Sharp' for Downsize works very well I got that tip from someone very smart

[Zom-B]

I find using 'Radial' for Splat and 'Sharp' for Downsize works very well I got that tip from someone very smart

I only can support that, give it a try and you'll see, the sharp downsizer is awesome

[fused]

I find using 'Radial' for Splat and 'Sharp' for Downsize works very well I got that tip from someone very smart

I only can support that, give it a try and you'll see, the sharp downsizer is awesome

I would have made it the default in Cindigo, but the same smart person told me that it's not a "best in all cases" solution.

It usually works great tho

[zeitmeister]

Alright, gonna give it a try...
thank you so far!
Nevertheless, a little more detailed information about the Mitchell-Netravali settings would be nice.

[Zom-B]

I would have made it the default in Cindigo, but the same smart person told me that it's not a "best in all cases" solution.

It usually works great tho

Yes, the radial splatting Filter is quite blurry if you use no Super sampling at all!

[lycium]

The reason I suggest the current "safe" defaults over my preferred Radial+Sharp is so that when someone renders an un-tonemapped EXR, they don't get nasty black rings around their highlights. For normal screen display, I really like that R+S combination (with SS 2+ as Zom-B says) - produces very nice images!

On the other hand, zeitmeister is looking for finer details, so maybe something else is appropriate. Does Cindigo let you adjust the filter radius and MN ring+blur parameters?

[Zom-B]

On the other hand, zeitmeister is looking for finer details, so maybe something else is appropriate. Does Cindigo let you adjust the filter radius and MN ring+blur parameters?

yes

[lycium]

If you want really sharp images, I can suggest the Catmull-Rom filter, which can be created from MN using the settings:

Blur = 0
Ring = 0.5
Radius = 2

If you're really after sharpness, try Ring = 1.0. You can still use the Radial splat with this downsize, which I'd suggest to do. For more discussion of the MN filter, you can see the classic paper here: http://www.cs.utexas.edu/~fussell/cours ... tchell.pdf