Fundamental Questions based on Brian Smith articles

[DF02]

Hi All,

I just download and read Brian Smith article called Critical VRay Settings Part 1 and have some basic / fundamental questions:

About Irradiance Map Min/Max Rates
I understand that:
Min 0 Max 0 Bucket 64x64 will render at resolution 64x64 IrMap which is same size of the final rendered image.
Min -1 Max -2 Bucket 64x64 will render at resolution 32x32 IrMap which is half size of the final image.
Min -2 Max -2 Bucket 64x64 will render at resolution 16x16 IrMap which is quarter size of the final image.
Question is:
If we set Min 1 Max 1 Bucket 64x64, what is the resolution of IrMap that it will render? It can't render 128x128 on 64x64 pixel image, can it?

About Adaptive Subdivision Smapler Min/Max Rates
I understand that:
Rate 0 means 1 sample per pixel.
Rate 1 means 4 samples per pixel.
Question is:
Rate 2 means 9 samples per pixel? or 16 samples per pixel?

[studioDIM]

it's 2 to the power of N
0 is 1
1 is 2 (not 4)
2 is 4 and so on

Lele

[DF02]

StudioDIM (Lele),

Thank you very much for the reply, but I got more confused.

When you said
it's 2 to the power of N
0 is 1
1 is 2 (not 4)
2 is 4 and so on
Does it mean
0 is 1x1=1 Sample
1 is 2x2=4 Samples
2 is 4x4=16 Samples
3 is 8x8=64 Samples
and so on?

In Brian Smith articles, it's written like this:
A rate of 0 means one sample per pixel.
A rate of 1 means 4 samples per pixel.

If you really means 0 means 1 sample per pixel, and 1 means 2 sample per pixel, then the question will be
"How you divide one pixel into 2?" Means, is it divided into 2 vertically or horizontally? Because to my understanding is the division must be equal vertically and horizontally.

Really appreciate your help, but perhaps can enlighten me further or Vlado perhaps?

[Sushidelic]

for the adaptive subdivision image sampler,according to the manual this is not quite correct. You could rather think of it as the numer of pixel edge subdivisions, which would lead to the formula
(n+1)^n as the actual number of samples taken.
0 = 1
1 = 4
2 = 9
3 = 16
tbc
Vlado states, that the max/min rates in the irradiance map are similar to those in the adaptive subdivision sampler, so the math for the resolution should be the same.

About Irradiance Map Min/Max Rates
I understand that:
Min 0 Max 0 Bucket 64x64 will render at resolution 64x64 IrMap which is same size of the final rendered image.
Min -1 Max -2 Bucket 64x64 will render at resolution 32x32 IrMap which is half size of the final image.
Min -2 Max -2 Bucket 64x64 will render at resolution 16x16 IrMap which is quarter size of the final image.
Question is:
If we set Min 1 Max 1 Bucket 64x64, what is the resolution of IrMap that it will render? It can't render 128x128 on 64x64 pixel image, can it?

--> min0 max0 = 64x64
--> min-1 max -2 should be min-2 max-1 and equals an irradiance map between 32x32 and 16x16 depending on the quality values you set (adaptive!!).
--> min1 max1 will render an irradiance map that is 128x128.
you can think of the resolution as a grid mapped over your screenspace, that can have for sure more subdivisions than your rendering does.

Some more reading on this:

http://www.spot3d.com/vray/help/150R...agesampler.htm
http://www.spot3d.com/vray/help/150R...vancedimap.htm

If you really means 0 means 1 sample per pixel, and 1 means 2 sample per pixel, then the question will be
"How you divide one pixel into 2?" Means, is it divided into 2 vertically or horizontally? Because to my understanding is the division must be equal vertically and horizontally.

So now you might think that your world is fine again, as all our pixels get sliced symmetrically through the middle like a cake, like you stated above.
But things are different for the qmc based stuff, where an arbitrary number of samples can be taken for a pixel, might it be 1,3,17 or 97.

As we're in all cases mentioned above often gather more than 1 sample per pixel, this is called "supersampling". So how to get a pixel value out of those? - This is, where this link might be handy:
http://en.wikipedia.org/wiki/Supersampling
This goes even further if you start digging into antialiasing filters like Area, catmull rom or similar. Thats some sweet thing, feel free to google!

regards,
michael

[DF02]

Re: Fundamental Questions based on Brian Smith articles

Thanks alot, Michael!! That's really helps!

As for this item:

Min -1 Max -2 Bucket 64x64 will render at resolution 32x32 IrMap which is

What I mean is actually -1/-1, not -1/-2 ... it was typo-error.

Thanks again!!

[studioDIM]

it's 2 to the power of N
0 is 1
1 is 2 (not 4)
2 is 4 and so on

Lele

Indeed, i was wrong there.
Glossed over the fact that you spoke of the adaptive subdivision sampler, rather than the Adaptive QMC sampler (which I verified when preparing the PPT calculator script to be 2^n, provided both share the same oversampling method, that is).
However, i got confused myself now reading the manual on spot3D:

Max. rate - controls maximum number of samples per pixel; zero means one sample per pixel, 1 means four samples, 2 means eight samples etc.

What is the formula for these numbers?
And how do they tie up to the ones described above?

Thanks to anyone which cares to elaborate.

Lele

[Sushidelic]

uuh damn, i was confused by the manual too!

basically, its simply 2^(n+1)

sorry!
michael

EDIT: ok now i'm completely confused too... as this is bulls**t above...
Vlado, could you please enlighten us?

[Sushidelic]

so quickly for fun:

logic rows would be:

0=1
1=2
2=4
3=8
4=16

or

0=1
1=4
2=9
3=16
4=25

else?

[studioDIM]

Next in line it's got to be the fibonacci sequence, rofl!

Lele

[apexmotion]

for the adaptive subdivision image sampler,according to the manual this is not quite correct. You could rather think of it as the numer of pixel edge subdivisions, which would lead to the formula
(n+1)^n as the actual number of samples taken.
0 = 1
1 = 4
2 = 9
3 = 16

It's actually (n+1)^2

[studioDIM]

if 2 means 8 samples, the one above is incorrect.
(2+1)^2 makes 9, not 8...

Fibonacci sequence wins, or the manual has a typo

edit
2^(n+1) explains both the values for 1 and 2 subdivs:
2^(1+1)=4
2^(2+1)=8
2^(3+1)=16

but it would only mean that there's some sort of "switch" if sampling is 0, or 2^(0+1) would lead to two, whereas it leads to 1 sample per pixel...

Lele

[DF02]

Scratching my head ...
Will it leads to the holy grail too? ... hahahaha ... just kidding


Looking for answer from Vlado ...

[studioDIM]

LOL!
It may just go like 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55...
There'd be loads of rabbits around, by then, though, don't raise sampling too much!

Lele

[Candelero]

LOL!
It may just go like 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55...
There'd be loads of rabbits around, by then, though, don't raise sampling too much!

Lele

Oh, you know about the rabbits... hehehe
BTW, this row also describes the seeds of a pine-cone. It is the same way the spiral goes from top to the bottom... 1,1,2,3,5,8...

Best regards,
nikki Candelero

[studioDIM]

Lol Nikki, yes!
And the sunflower seeds, and many other natural phenomena.
There's a gorgeous book on the mathematics of the living systems in nature, warmly suggested to the ones with a curious mind: no mathematics is actually involved (well, very very little), and i found it very readable and just that bit too short

http://plus.maths.org/issue19/reviews/book1/index.html
(the version i read, bought in the Uk)
http://www.amazon.com/Lifes-Other-Se...ion/0471296511
(a paperback, likely published in the US)

Lele