Post a reply to this message

Attachments:
Download 'rad1.jpg' (31 KB) Download 'rad2.jpg' (27 KB) Download 'rad3.jpg' (32 KB)

Preview of image 'rad1.jpg'

Preview of image 'rad2.jpg'

Preview of image 'rad3.jpg'

This is the first time I see those incredibly soft shadows some commercial
renderers produce (they don't look always realistic, though) done in
POV-Ray. Did you use pure radiosity lighting or in a combination with area
lights? How long did it take to render these pictures? And the most
important question: Which color has that light? I am always trying to get
the right shade of yellow/orange for my lights ;)

Post a reply to this message

Lukas Winter wrote:
> Did you use pure radiosity lighting or in a combination with area
> lights?

  I used area lights, of course.

> How long did it take to render these pictures?

  Between 30 minutes and 1 hour.

> And the most
> important question: Which color has that light?

  White.

Post a reply to this message

Nice images.

I wrote a little program to combine the three images dynamically 
depending upon the mouse position on the screen.  It uses a very simple 
and not quite correct method but the results are interesting.

Sort of simulates varying the intensity of any of the three lights.

Unzip to a directory and run the exe.

Post a reply to this message

Attachments:
Download 'imagemerge.zip' (701 KB)

Lukas Winter <web### [at] removeitgeloeschtnet> wrote:
> This is the first time I see those incredibly soft shadows some commercial
> renderers produce... done in POV-Ray.

I see you're new to povray. :)

Povray has got area light soft shadows for even longer than radiosity...

Post a reply to this message

Paul Fuller <pgf### [at] optusnetcomau> wrote:
> Nice images.
>
> I wrote a little program to combine the three images dynamically
> depending upon the mouse position on the screen.  It uses a very simple
> and not quite correct method but the results are interesting.
>
> Sort of simulates varying the intensity of any of the three lights.
>
> Unzip to a directory and run the exe.

That is interesting, I wonder how far you could push this trick?

I'm thinking out loud here, but bear with me. Lets say each image was 50KB,
and you loaded the images into the fast RAM on a video card. With 100MBs of
RAM you could have 2000 total images in memory, or alternatively your scene
rendered with the lightsource placed at each point of a 45 by 45 grid. You
would only need to sample the three or four images with the closest
lightsource to simulate light positions between grid points, so it seems
like computationally it would be feasible. With the use of an adaptive grid
you could simulate real-time radiosity with all the bells and whistles in
extrememly complex scenes.

So... any chance you could share your code?

Post a reply to this message

scam wrote:
> Paul Fuller <pgf### [at] optusnetcomau> wrote:
>> Nice images.
>>
>> I wrote a little program to combine the three images dynamically
>> depending upon the mouse position on the screen.  It uses a very simple
>> and not quite correct method but the results are interesting.
>>
>> Sort of simulates varying the intensity of any of the three lights.
>>
>> Unzip to a directory and run the exe.
> 
> That is interesting, I wonder how far you could push this trick?
> 
> I'm thinking out loud here, but bear with me. Lets say each image was 50KB,
> and you loaded the images into the fast RAM on a video card. With 100MBs of
> RAM you could have 2000 total images in memory, or alternatively your scene
> rendered with the lightsource placed at each point of a 45 by 45 grid. You
> would only need to sample the three or four images with the closest
> lightsource to simulate light positions between grid points, so it seems
> like computationally it would be feasible. With the use of an adaptive grid
> you could simulate real-time radiosity with all the bells and whistles in
> extrememly complex scenes.
> 
> So... any chance you could share your code?
> 
> 
> 

There are definitely some possibilities for giving games a ray-traced 
like appearance without actually calculating everything dynamically. 
You are limited though by the combinatoric explosion that happens if you 
want to support too much with this method.

Probably better applied to local effects rather than the illumination of 
the whole scene as it is here.

Another trick / trade-off not done here is to render the basic scene 
then calculate each effect independently.  Calculate the difference 
between the original and new scene.  You don' necessarily need to store 
every different pixel.  Perhaps only store the significant changes or 
those within some subregion that is most significant.

By the way, there is a fault in the method as shown below.  The first 
image is displayed with Alpha=1.0.  Then the second image is drawn over 
it with an Alpha that depends upon how strong it should be.  The third 
image is again drawn over that and I think has too strong an influence 
on the result.  There needs to be some adjustment of the Alpha from the 
simple method used here.  Perhaps it could all be made additive...

I wrote the program in BlitzMax which is great for quick graphics stuff. 
  the graphics routines support hardware accelerated alpha and rotation 
which is nice.  You may not have or know of BlitzMax but the code is 
pretty straightforward.

Also it is just knocked together after seeing Warp's post.  Probably too 
many globals and other frowned upon practices for the purists.

Anyway, publish and be damned:

SuperStrict

Global G_ScreenWidth:Int = 800
Global G_ScreenHeight:Int = 600
Global G_MaxDistance:Float = G_ScreenHeight * 2 / Sqr(3)

Graphics G_ScreenWidth, G_ScreenHeight,1

SetBlend(ALPHABLEND)

' H:\Resources\Images\System\QiD Logo 1920x1200.png
' H:\Resources\Images\System\QiD Logo 1920x1200 - Gold.png

Global G_Image1:TImage = GetImage("Warp - Rad1.jpg")
Global G_Image2:TImage = GetImage("Warp - Rad2.jpg")
Global G_Image3:TImage = GetImage("Warp - Rad3.jpg")

Global G_P3:TPoint = TPoint.CreatePoint(G_ScreenWidth / 2, 0)
Global G_P1:TPoint = TPoint.CreatePoint(G_P3.X - G_MaxDistance / 2, 
G_ScreenHeight - 1)
Global G_P2:TPoint = TPoint.CreatePoint(G_P3.X + G_MaxDistance / 2, 
G_ScreenHeight - 1)

Global G_Mouse:TPoint = Tpoint.CreatePoint(0.0, 0.0)

Global G_Offset:TPoint = TPoint.CreatePoint((G_ScreenWidth - 
ImageWidth(G_Image1)) / 2.0, (G_ScreenHeight - ImageHeight(G_Image1)) / 2.0)

While Not KeyHit(KEY_ESCAPE)

	Cls

	G_Mouse.SetXY(MouseX(), MouseY())
	
	Local S1:Float = CalcStrength(G_Mouse, G_P1)
	Local S2:Float = CalcStrength(G_Mouse, G_P2)
	Local S3:Float = CalcStrength(G_Mouse, G_P3)
	
	DrawImageFade(S1, G_Image1)
	DrawImageFade(S2, G_Image2)

	If (G_Image3 <> Null)
		DrawImageFade(S3, G_Image3)
	EndIf
	
	DrawMarkers()
	
	Flip
	
Wend

Function GetImage:TImage(filename:String)

	Return LoadImage(filename, FILTEREDIMAGE | MASKEDIMAGE)
	
End Function

Function CalcStrength:Float(p1:TPoint, p2:TPoint)

	Local dx:Float = p1.X - p2.X
	Local dy:Float = p1.Y - p2.Y
	
	Local distance:Float = Sqr(dx * dx + dy * dy)
	
	If (distance > G_MaxDistance)
		Return 0.0
	Else
		Return 1.0 - distance / G_MaxDistance
	EndIf

End Function

Function DrawImageFade(s:Float, image:TImage)

	SetAlpha(s)
	DrawImage(image, G_Offset.X, G_Offset.Y)
	
End Function

Function DrawMarkers()

	SetAlpha(1.0)
	SetColor(255, 255, 255)
	
	DrawMarker(G_P1)
	DrawMarker(G_P2)
	DrawMarker(G_P3)

End Function

Function DrawMarker(p:TPoint)

	DrawOval(p.X - 1, p.Y - 1, 3, 3)
	
End Function

Type TPoint
	Field	X:Float
	Field	Y:Float
	
	Function CreatePoint:TPoint(x:Float, y:Float)
	
		Local T:TPoint = New TPoint
		T.X = x
		T.Y = y
		
		Return T
	
	End Function
	
	Method SetXY(x:Float, y:Float)
	
		self.X = x
		self.Y = y
		
	End Method
	
End Type

Post a reply to this message

Paul Fuller <pgf### [at] optusnetcomau> wrote:
>
> There are definitely some possibilities for giving games a ray-traced
> like appearance without actually calculating everything dynamically.
> You are limited though by the combinatoric explosion that happens if you
> want to support too much with this method.

Thanks for posting the code Paul, I hadn't heard of BlitzMax before but it
looks like it will be fun to play with. They even offer a trial of the
software, so I've been able to compile your code myself.

Even with the various limits I think you could still use your approach to
make a cool little platformer or something. Each level could be a single
window, and the protagonist could be exploring a dark world with an area
light source (like a lantern or a burning torch). With some focal blur,
area lights and radiosity I think you could create a really cool, unique
visual style.

Post a reply to this message

Am Thu, 03 May 2007 00:01:07 -0400 schrieb nemesis:

> Lukas Winter <web### [at] removeitgeloeschtnet> wrote:
>> This is the first time I see those incredibly soft shadows some
>> commercial renderers produce... done in POV-Ray.
> 
> I see you're new to povray. :)
> 
> Povray has got area light soft shadows for even longer than radiosity...

I have been using POV-Ray for about two years now, although not very
often. I know there are area lights but they don't have the effect I
thought of. Perhaps I should have said soft shading. I have the impression
most POV-Ray images have very "sharp" global illumination, everything
looks a bit too exact to be believable. Maybe it's the lack
of a light bleeding filter on the final image. As someone noticed, area
lights also only have an effect on shadows, not on highlights or diffuse
reflection.

Post a reply to this message

Am Thu, 03 May 2007 00:28:18 +0300 schrieb Warp:

> Lukas Winter wrote:
>> Did you use pure radiosity lighting or in a combination with area
>> lights?
> 
>   I used area lights, of course.
> 
>> How long did it take to render these pictures?
> 
>   Between 30 minutes and 1 hour.
> 
>> And the most
>> important question: Which color has that light?
> 
>   White.

Really? It looks a bit yellow.

Post a reply to this message