color filter for camera

[rossgilmour]

I want to apply a color filter to a camera that only lets through the color I
require. I can do this when the 3D member is on the stage using a solid color
in a sprite on top of the sprite containing the 3D member and setting the Ink
to transparent. However I want to make this a property of the camera. Any
suggestions?
Ross

[Newt99]

1. Create a new bitmap member by code, draw a filled rectangle in it and set
its color 2. Create a texture out of this bitmap member 3. Set the new texture
as an overlay of the camera Cheers, Karl. [url]http://www.chromelib.com[/url]

[rossgilmour]

Karl,
Many thanks but when I tried this all I see is solid color.
Ross

[Newt99]

Ross,

Step 4 was missing... here it is:

4. Set the overlay blend to 50 or so

Cheers,
Karl.

[url]http://www.chromelib.com[/url]

[rossgilmour]

Karl,
Unfortunately this only gives a colored tint to everything. I want it to act
as a color filter that does not let through certain colors. For example if I
want a red filter then anything that is black, blue or green will be black
while anything that is white, red, yellow, magenta ... will be shades of red
depending on the red value of the RGB color.
Ross

[nick kang]

Sounds like you want to do image processing in real-time. That's probably only
possible using a 3D engine that supports DirectX9 and a custom (but simple)
pixel shader. With SW3D, you might get try using getPixel() and setPixel() to
process the current frame and then displaying the result. Depending on the
size of the 3D viewport, it might not take too long. Also keep mind that the
image data of a 3D member is fixed at 320x240, so if you need higher than that,
you'll have to turn directToStage off on your 3D sprite, then copyPixel() the
stage to your image buffer, then do your pixel processing on that.
--------------------------------- nick * rasterwerks - com

[rossgilmour]

Nick,
Yes I do want to do image processing real-time. I am working on a displaying
stereographic views of 3D environments as anaglypic images which uses the
red/cyan glasses. To do this I need to produce 2 seperate images (one with only
the red channel and the other with the green & blue channels) and then add the
2 images together on the stage. I have already done it by setting 2 bitmaps to
different camera positions and using copyPixel with an Ink to remove the
unneeded colors. This works but is a bit slow. You can see an example of this I
have done if you have a pair of red/cyan glasses (red/green or red/blue also
work) [url]http://www.rg-media.co.uk/stars3D/stars3D.htm[/url]

I was hoping to find a faster method.
Many thanks,
Ross

[jrPace]

I'm talking off the top of my head right now, but is it possible to set the
blending mode of the overlay to #multiply ? Or is only #blend supported? If
I've got this right in my head, mulitplying the image in the scene by a purley
red (255,0,0) image, *should* give you what you want. ie say one of the pixels
underneath is (128,64,32). If you multiply this by (255,0,0) then you get
(after normalisation) 0.5 * 1 0.25 * 0 0.125 * 0 = (0.5,0,0) or
(128,0,0) Or I could be talking out of a certain orifice :) Jonathan Pace
[url]www.unwrong.com[/url]

[nick kang]

I see, for an anaglyph you need the overlapping renders of a particular model
to be additive, like in your demo. Unfortunately, there's no additive blending
between fills in SW3D, so you'll need to use imaging functions after all. Too
bad, you can almost get away with just using multiple cameras if a wireframe
display is all you needed (because image overlap would be minimal), then it
would run full speed. --------------------------------- nick * rasterwerks -
com

[jrPace]

But there *is* additive blending between texture stages. So you could render
each view (red, green, whatever) into a texture, then layer up these textures
onto a quad that fills the screen using the #add blend mode. It's getting
pretty convoluted, but I think there's a way here to get it working without
resorting to imaging lingo. Jonathan Pace [url]www.unwrong.com[/url]

[nick kang]

Good point, but unfortunately there isn't a direct way to render to a texture
(and thus keeping all the operations on the video card). You still need to
access the member3D.image or by doing a copyPixel of the stage, then create a
new texture from that image, which will incur a similar performance hit. But
maybe you have another method in mind? I slapped together a test, and the
overlap is problematic. Looks okay in wireframe (my glasses have the red
filter over the left eye), not good filled, even with blending. Enable shading
and it looks like a mess. [url]www.rasterwerks.com/link/anaglyph_001.htm[/url]
[url]www.rasterwerks.com/link/anaglyph_001.dir.zip[/url] (src)
--------------------------------- nick * rasterwerks - com