Xabre Supersampling
- Thread starter parhelia
- Start date
Pete said:Mmmm, bilinear filtering.
And 2x does look like a blur filter. 3x and 4x looks decent. I notice there are no framerates in the shots.
3x and 4x looks like ordinary 2xOGSS, not very good in other words
the performance drop is very steep though, "4x" really has the drop I'd expect from 4x
OpenGL guy
Veteran
2x looks like it was rendered at 1/4 resolution and stretched to fullscreen. It's awful.Pete said:Mmmm, bilinear filtering.
And 2x does look like a blur filter. 3x and 4x looks decent. I notice there are no framerates in the shots.
I really can't see any AA in the 3x or 4x shots.
Well, my screen res isn't sufficient to look at the whole picture at once--I was concentrating on how the textures on the floor and the grates on the left look sharper in 3x and 4x. Looking again, I can't see significant AA on the gun or the left and right "sidewalks."
My sarcasm stands.
My sarcasm stands.
Assuming that I've got the information correct and haven't misunderstood what's be said...
The only way is to pre-filter the theoretical image, however that is difficult to do and so the 'next best thing' is to sample at a higher rate. That moves the aliasing cut-off frequency point up the spectrum and, as the high frequency content usually drops off with increasing frequency, the visible aliasing decreases.
In summary, you can't make a "silk purse out of a sow's ear". There were complaints that the Nvidia Quincunx 2x filter was blurring, but that, at least least, was a genuine 2x supersample.
No it can't. Once the aliases, i.e. high frequency information that was in the 'theoretical' image prior to sampling but is now masquerading as low frequency data, are in in the image, it is not possible to remove them without damaging the valid low-frequency data. All a low-pass filter does is implement a digital version of a very old "hardware" technique. that was applied post-RAMDACNagorak said:
The only way is to pre-filter the theoretical image, however that is difficult to do and so the 'next best thing' is to sample at a higher rate. That moves the aliasing cut-off frequency point up the spectrum and, as the high frequency content usually drops off with increasing frequency, the visible aliasing decreases.
In summary, you can't make a "silk purse out of a sow's ear". There were complaints that the Nvidia Quincunx 2x filter was blurring, but that, at least least, was a genuine 2x supersample.
2xRGMS + 5-tap gaussian if my memory serves me well.
Recent updates on XabreII:
http://xabregamers.dgwh.com/modules/nsections/index.php?op=viewarticle&artid=13
