Eye space vs screen space

Graphics and GPU Programming Programming

Started by CyberRascal March 12, 2012 02:06 PM

2 comments, last by ATEFred 12 years, 1 month ago

208

Author

March 12, 2012 02:06 PM

Hey,

I'm a bit worried this isn't the correct place to post this, but here goes:

What is the difference, if any, between eye space and screen space?

I've sort of made the mental assumption that eye space is the 'world space' represented in an ON-basis placed in the 'viewport' plane, with the normal of that plane as the third vector.

In contrast, in my mind, screen space is spanned by the plane of the viewport, with geometry projected onto the screen plane. So, screen space is no longer 3D, but a 2D representation ready for rasterization / shading and stuff.

Does anyone care to tell me I'm stupid? Or pleasantly correct me? Or tell me I'm kind of right?

Hodgman

52,717

March 12, 2012 02:09 PM

View / eye / camera / screen - space are sometimes used interchangeably in a lot of literature, which confuses things...
Though yes, I usually use 'screen-space' when I'm referring to a 2D coordinate system --- either in integer pixel coordinates ([0,0] to [width-1,height-1]), normalized pixel coordinates ([0,0] to [1,1]), or NDC ([-1,-1] to [1,1]).

If someone said 'screen-space' when they actually meant 3D view-space, I'd get a bit confused.

View / eye / camera - space are all the same to me, consisting of the 3D world-space points that have been transformed by the view matrix (inverse camera matrix).

Post-projection space / NDC-space are the 3D view-space points that have been transformed by the projection matrix and have undergone perspective division. From here you simply drop the z component to get screen-space.

Though, I've also seen people refer to post-projection-space to be the 3D view-space points that have been transformed by the projection matrix but have not yet undergone perspective division... And once that division occurs, you're in NDC-space (screen-space with z).

. 22 Racing Series .

CyberRascal

208

Author

March 12, 2012 03:29 PM

Thanks, that clears it up.

Another question:

http://www.opengl.org/sdk/docs/tutorials/ClockworkCoders/loading.php

The above page says:

"A fragment is basically a pixel before it is rasterized."

This confuses me a bit. I thought rasterization was the operation vector -> dot / pixel. But I've learned that a fragment is basically a 'possible' pixel for a certain location, that can be blended together with other 'possible' pixels in the same location, which means that a fragment is actually something that is rasterized, which doesn't seem to mesh with that quote from opengl.org. Can somebody clear this up?

ATEFred

1,700

March 12, 2012 06:00 PM

fragment vs pixel is kind of inconsequential. As far as I understand, a in OGL pixel shaders are called fragment shaders because a fragment does not automatically correspond to a pixel. For example, in the pixel/fragment shader you could do a texkill, which would result in no pixel being rasterized. I could be wrong though, never spent too much time mulling it over...

Eye space vs screen space

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Eye space vs screen space

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines