Quake3World

**

Just wanted to say that this thread is very, very cool. Pictures always do a hugely better job of getting stuff like this across than words do.

Excellent work.

[Edit: Just for clarity, you might want to add something in the first post about what version of Q3map2 you're using, and maybe a list of what switches you plan on investigating? This thread might be good as a sticky or some sort of saved-off resource when you're finished.]

Yeah, keeping an eye on this thread... it's going to be added to the sticky "archive" thread.

**

The argument for the -filter parameter is supposed to be measured in world units IIRC. So you won't see much softening in the luxels (lightmap pixels) until you extend the radius far enough to encompass at least the neighbouring luxel. This seems to be pretty consistent with your findings.

I think the gaussian blur comparison is probably spurious - it would be worth trying the same test with higher values to confirm or deny it.

Shallow wrote:The argument for the -filter parameter is supposed to be measured in world units IIRC.

So what matters is how much the part of the lightmap which is mapped to a face is stretched.
Becuase in most cases it is stretched, it makes sense that unit of this value is a in-game unit and it is the same unit with texel, right?
At the same time, now I see my experiment just above was out of point.
And conclusions derived from all these smoothing tests might contain errors to some extent depending on the way lightmaps are mapped.
Thank you.

a13n wrote:There do exists a difference as PIC04 shows. But it is subtle and ignorable. Therefore I conclude that facing angle of a brush does not affect bouncing.

Well, you shouldn't see a drastic difference since we are looking at radiosity, so after a number of traces, the lightmap gets lit up pretty much the same (you're bouncing multiple times, so any differences are masked by the repeated bouncing). But Q3Map2 does take surface angle into consideration. Your example doesn't show it very well, but I remember my lightmap experiment showed how light bounced between two blocks. It's more pronounced with shadows than well lit areas.

a13n wrote:-dirty switch dirtifies lightmap with black ink around brush boundaries. Q3map2 readme simply says "ambient occlusion" and now it makes sense.

Not really. It doesn't just paint black areas around every brush boundary. Ambient occlusion is an lighting method that simulates raytracing by taking into account other obstructive geometry. It actually calculates distance between the obstructive geometry and then determines the relative darkness from that. Think a crevice, it should be much darker inside a crevice than it is outside since they are so close together, very little light can get in between the cracks.

Clarification of terms:
Pixel - actual dots on your screen
Texel - texture pixel
Luxel - lightmap pixel
Makes it easier to understand if you differentiate between them.

-filter works between two different scales. It is measured in a radius of game units, but lightmaps are luxels that can be stretched to any scale relative to game units. So reducing lightmap resolution requires an increase in -filter value (increased game unit radius) to compensate for larger luxels.

-samples and -super works in a similar way, but -samples uses a faster, but slightly less accurate algoritym. IIRC, ydnar said that there was little noticable difference, so you should generally just use -samples. -samples increases the light attenuation, while -filter adds a gaussian blur to the lightmap. In terms of quality, -samples > -filter

@obsidian
Thank you for your comments.
I learned something new.
But just remember here is a screenshots thread.
Your explanation lacks them.

Bounce 4 probably reaches to the end of the tunnel but with such a low illuminance that you can't see it. It's diffuse light and walls in radiant, not a laser and mirrors.

a13n wrote:But just remember here is a screenshots thread. Your explanation lacks them.

Tis called teamwork, matey... you post the screenshots, I comment on them!

Teamwerk?

a13n wrote:@obsidian
Thank you for your comments.
I learned something new.
But just remember here is a screenshots thread.
Your explanation lacks them.

Chill out. There's nothing wrong with discussion.

If there are two shader-manipulated textures like below and there is an entry on shaderlist.txt like below, selecting textures->subdir in radiant will load both textures.
This seems to be the same principle with the post above though this might not be practical at all.

An excel-gen texture exported from q3ase.


EDIT:
I've never noticed that q3ase can export each frame till today, which gives continuous photoshop filtered-like effects effectively. :icon28: