Photo Maps

[Crazy Homeless Guy]

You can save and reuse the photon map independent of geometry and image size; doing so can dramatically speed up rendering for a camera walkthrough. Here we save it to try different lookup options to resolve the low-frequency noise. We run once with 2.5 million photons shot from the light source and store it in the photon map file.

 

Can anyone help me decipher this statement? ...it was in the 3dsmax help files for Mental Ray. Specifically the statement of re-using the photon map independent of geometry and image size?

 

When I read this, it sounds to me that if I change the geometry, the Photon Map is still valid. Which doesn't make a lot of sense when I try to think about it.

[muchhal.sandeep@gmail.com]

Hi Travis,

 

I shall say that we both are feeling the same.

as we change the geometry we shall recalculate the PMAP.

 

the image size has got no problems

once we save the photon map then we can render it at any required size need not recalculate.

[Crazy Homeless Guy]

good to know. i was reading a another mental ray thread post today, and it referenced a different anomaly in the help files max mental ray help files.

[Crazy Homeless Guy]

I assume that the help is making a distinction between possible

GI solutions that people may be using, such as Radiosity which

is very much Geometry dependant. Chabnge the geometry and

you will need to recalc your GI solution, which is not the case

with photon mapping. (as a rule)

 

regards

Bri

 

Why would you not need to recaculate the photon mapping if you change the geometry? ....this is what doesn't make sense to me. maybe i need to study more on how photon mapping works.

 

The majority of my GI experience is with Vray's irradiance map, so I am processing this in relation to those terms. Using irradiance passes, the map is representation of the 3d geometry in the scene. If I change the geometry, then the map created by the irradiance solution does not have the proper geometry to align to, and becomes invalid.

 

It is this thinking that is confusing me on photon mapping. I remember the tutorial saying something about the way it is calculated, but I can't remember exactly what it is now. I will look at it again in the morning, and try to decipher it.

[Crazy Homeless Guy]

It works with any light, no matter how "unphysical,” since it is concerned only with the transport of light from surface to surface, not the transport of light from the light to the surface.

 

So. What I am making out is that the photon map is a map of the light, or more precisely the transport of the light. This is why changing the geometry does not affect the map. The photon map already has the amount of light that will strike a newly introduced surface or geometry modification stored into the file. However, if I add additional, or change the light source, I will need to re-compute the photon map because the map of the light is no longer accurate to the actual lights in the scene?

 

Whereas, or at least I believe that, irradiance maps pretty much become invalid as soon as the geometry is changed. I say this because if you open the Vray irradiance map file in the irradiance map viewer, you can see that that the surfaces are defined by how the light hits them. If you move a surface, the saved irradiance map does not reflect that move without recalculating. Or at least it does not reflect it with any deal of accuracy.

 

Is there a viewer, or any way to look at a photon map to better understand how it works?

 

…also, I assume that placing geometry close to a light source will affect how the photon map should behave, therefore making the saved map un-accurate because the distribution of light in the scene would have changed(?).

[Crazy Homeless Guy]

after more research, i have to go with my initial inclination. ...when you change the geometry in the scene, you should recaculculate your photon map, because it is not accurate.

[iduna79]

Well, photons bounces against the geometry. If the geometry is moving you have to recalculate the photon maps (the same goes for fg maps) or the calculations will be incorrect. Otherwise the saved photons bounces against something not there and so on.

[MasterZap]

I'd say this snippet of manual text is suboptimally worded.

 

The photon map is "geometry independent" in exactly the way Brian explains, i.e. it actually stores it in a point-cloud data structure in space, not actually "on" the geometry *itself* (even though the points *are* "on" the geometry).

 

So, changing, say, the number of segments in a sphere changes nothing worthwhile to the photons, whereas it would completely break the traditional "Radiosity" method.

 

It's probably this distinction the manual is trying to (poorly) convey. I.e. back in the radiosity days, the data was intrinsically linked to vert's and tri's, so you couldn't even do the most trivial little thing or you *completely* broke the entire radiosity solution.

 

/Z