Post Process and world transforms

Borgleader

Hey guys,

I was just wondering if there was any way in a Post Process to compute the position in 3D space of each pixel?

My first thought was to take the Camera position and add to it the Camera vector multiplied by the value in the depth buffer. But I soon realized that by that stage of the rendering process the camera is most likely at (0,0,0) and facing -y.

Any ideas? Thanks.

imbueFX

I believe you're looking for the World Position expression.

(edit: in the material editor, to be specific)

mAlkAv!An

May I ask for what kind of post process effect you need this?
Not so long ago I did reconstruct the view space position from depth in a pp material, world space might be doable as well.

Borgleader

http://www.polycount.com/forum/showthread.php?t=97548
It's not exactly this, but the idea is similar. I essentially need to know during the post process stage is the pixel is near the player/pawn (which is a position i send to the shader in the form of 3 floats).

Froyok

I don't think you can access world data in a post-process material.

imbueFX

Ah I didn't know that Froyok.

If it's a 1st person camera, could you just used pixel depth instead?

Rekmar

while you can't access world data in a post process material, you can probably re-construct it with some hard work. I myself re-constructed a "view-space" position (position relative to the camera, ie. Z is always depth)
the basis is the scene depth, then I used a screen-covering horizontal gradient for X and a vertical one for Y. I figure if you can turn those gradients "3 dimensional" and "rotate them accordingly", and feed the shader with the camera's location and rotation, you might get it done

Borgleader

Thanks for the suggestion Rekmar, I'll take a look into it.