Eye position in post-processing shader

[Rastar]

Hi,

 

for a post effect I need to reconstruct the pixels' camera position from the depth buffer. I am a little confused since Leadwerks seems to be doing some things differently than textbook OpenGL, so a few questions I have:

 

I find this snippet e.g. in the directionallight.shader (and also a little different in klepto's sky shader)

 

screencoord = vec3(((gl_FragCoord.x/buffersize.x)-0.5) * 2.0 * (buffersize.x/buffersize.y),((-gl_FragCoord.y/buffersize.y)+0.5) * 2.0,depthToPosition(depth,camerarange));
 screencoord.x *= screencoord.z / camerazoom;
 screencoord.y *= -screencoord.z / camerazoom;
 screennormal = normalize(screencoord);
 if (!isbackbuffer) screencoord.y *= -1.0;

 

Now I get the depthToPosition function to calculate the z value. However, the x and y coordinates confuse me:

1. First of all, it seems that Leadwerks uses the y position in the backbuffer inverted compared to the front buffer, therefore those isbackbuffer -> * -1 lines (here and also in other shader parts)?
   
2. Is that also the reason that the y coordinate uses 0.5 - gl_FragCoord.y rather than gl_FragCoord.y - 0.5?
   
3. The first line basically seems to compute normalized device coordinates, but why is the x coordinate multiplied by buffersize.x/buffersize.y?
   
4. This code seems to assume a symmetric FOV with equal FOV values for x and y, correct?
   

[Josh]

1. OpenGL Y screen coordinates are the inverse of what you would expect.

2. It worked out that way. See #1.

3. The camera perspective is decided by the vertical space, not the horizontal. This works out better with wide and narrow screen resolutions.

4. Yes, sort of. I mean, if the screen width is wider than the height, which it almost always is, the width will have a wider field of view, but it will be proportional to the vertical FOV and screen height.

 

See the light shaders for pixel position reconstruction (although I think it is relative to the camera).