|
 |
On Tue, 22 Apr 2003, Christopher James Huff wrote:
>In article <Pine.GSO.4.53.0304221425230.5704@blastwave>,
> Dennis Clarke <dcl### [at] blastwave org> wrote:
>
>> is that a gaussian function that is applied to calculate the anti-aliased
>> value of the pixel or simply the average of pixels within a 10 pixel radius.
>> Then again, are we talking a square of 10x10 and not a radius of 10 pixels?
>
>This is antialiasing, not blur. We are supersampling a single pixel.
In what way is antialiasing within a single pixel different from blur of
a single pixel. They are essentially the same thing n'est pas? My thoughts
on this are that you can have a +r3 antialias approach in which a 3x3 grid
should be laid down within a single pixel and then the average of the RGB
values calculated at the center. The concept of blur would be the same yes?
Simply with a blur you would increase the resolution of the image by a factor
of 3 in each dimension and then calculate the average RGB values within a
central pixel from the surrounding 3x3 pixels. Is the real difference, if
any, that this is more like a digital convolution by using a comb filter
function with a 3x3 identity matrix ( all 1's ) and therefore the central
pixel value would be including its neighbors while anti-aliasing does not
include its neighbors at all? Is the sampling matrix that is used for the
average actually all 1's or is it 1/sqrt(2) on the corners and sqrt(2/3)
on the edges?
antialias matrix for 3x3 within a single pixel
+--------+--------+--------+
| | | |
| p | q | p | where p = 1/sqrt(2) and
| | | |
+--------+--------+--------+ q = sqrt(2/3)
| | | |
| q | 1 | q |
| | | |
+--------+--------+--------+
| | | |
| p | q | p |
| | | |
+--------+--------+--------+
I'm just thinking via my keyboard here ..
Dennis
Post a reply to this message
|
|
