Re: Triangle rasterizer with A-buffer antialiasing
- From: Just d' FAQs <nobody-here@xxxxxxx>
- Date: Sat, 12 Nov 2005 08:42:03 -0600
On Sat, 12 Nov 2005 13:29:54 GMT, "Dave Eberly"
<dNOSPAMeberly@xxxxxxxxxxxxxxx> wrote:
>I disagree.
>
>The image processing community allows multiple definitions of
>the term "pixel", including "point" (a sample) or "box" (formed
>by four samples). The same is said for the term "voxel". Some
>image processing algorithms are better formulated using one
>definition versus the other. For example, the Marching Squares
>algorithm for isocurve extraction from 2D images requires the
>"box" formulation. And sometimes the "box" formulation will
>use the samples as box centers rather than box vertices.
>
>One might even argue that "point" has its problems since a pixel
>on the screen is not a mathematical point. This gets you into the
>topics of "inner scale", "hyperacuity", and others in the realm of
>computer vision.
I am identifying a pixel with a point sample, which is the standard
definition in discussions of aliasing and antialiasing. Some authors
for some algorithms may find it convenient to make a distinction, but
I insist on the classic meaning, *especially* in this thread.
As Jim Kajiya pointed out in a SIGGRAPH paper on antialiasing fonts,
the *display* of a pixel on the screen involves reconstruction with a
point spread function that tends to be gaussian. Careful writers have
always discussed both antialiasing and reconstruction, have found both
to be essential, and have distinguished between the two. The effect
called "pixellation", where each sample is displayed as a box, is no
more nor less than hideous reconstruction.
Hyperacuity -- discriminating higher resolution than cone spacing can
explain -- is not well-understood even among vision researchers, and
certainly not by me. But one possible explanation is relevant to the
topic of this thread: if the visual system is able to integrate over
more than an instant of time, it can "jitter the samples" to achieve a
kind of supersampling.
.
- References:
- Triangle rasterizer with A-buffer antialiasing
- From: kevinppp68
- Re: Triangle rasterizer with A-buffer antialiasing
- From: Just d' FAQs
- Re: Triangle rasterizer with A-buffer antialiasing
- From: Dave Eberly
- Triangle rasterizer with A-buffer antialiasing
- Prev by Date: Re: Triangle rasterizer with A-buffer antialiasing
- Next by Date: Re: how to create random points with a minimum and maximum distance from the next?
- Previous by thread: Re: Triangle rasterizer with A-buffer antialiasing
- Next by thread: rotation problem
- Index(es):
Relevant Pages
|
