ISSUE-2317 (feGaussianBlur stdDeviation=0): Give different result for feGaussianBlur stdDeviation=0 [SVG Full 1.1]

ISSUE-2317 (feGaussianBlur stdDeviation=0): Give different result for feGaussianBlur stdDeviation=0 [SVG Full 1.1]

Raised by: Doug Schepers
On product: SVG Full 1.1

>From David Dailey <>:

The spec (1.1) says:
A value of zero disables the effect of the given filter primitive (i.e., the
result is a transparent black image).
If the attribute is not specified, then the effect is as if a value of 0 were

Opera and Firefox both handle this correctly (according to the spec). But it
seems to me the spec is a bit too fussy. It is written, I presume, to avoid a
problem of dividing by zero in the definition of the blur function:

H(x) = exp(-x2/ (2s2)) / sqrt(2* pi*s2)

But when you think of it this way: more standarddeviation means more blur. Zero
standard deviation should mean no blur.
The issue arose when I wanted to use script to append a blur filter onto an
object and allow a slider to control the amount of blur. Rather than removing
the filter attribute from the DOM I just wanted to change StdDev to zero. Alas,
while ASV knew what I wanted to do, Opera and FF followed the spec. I think this
attribute is only used in feGaussian blur, but even if it were used in other
contexts, I do think that authors will think zero means "no variability", not
"turn the image black and make it transparent"

Is this the sort of thing that should be raised here, in SVG IG, or with SVG WG,
or with a nontrivial subset of the three?

Reply from Doug Schepers <>:

Interestingly, I ran into this very issue this past weekend. I agree
with you... the effects of values should be intuitive, especially when
being iterated though via script (this is why I think arc segments
should still render when the start and end points are the same).

Illustrator, whence come the filters, also agrees with you, though
Inkscape follows the SVG spec. I think this is something that should be
changed, since there are differences in implementations. If you
wouldn't mind resending your message to www-svg, I'll remember to raise
it in discussion with the SVG WG.

Received on Friday, 2 April 2010 15:49:20 UTC