Which one to buy: Tamron SP 180mm F/3.5 Di Macro or Sigma ..

External Since: Jun 03, 2004 Posts: 917

In article <62ia8jF23nen1U1.DeleteThis@mid.individual.net>, Chris Malcolm
<cam.DeleteThis@holyrood.ed.ac.uk> wrote:

> However many pixels you have, there will always be the problem of how
> to handle detail as you approach the sampling frequency. Do you dump
> all possible aliasing artefacts at the cost of losing some accurate
> detail in cases where aliasing artefacts are not obtrusive? Do you go
> far all possible detail and live with the aliasing artefacts produced
> in some cases? Do you go somewhere between those two extremes?
>
> It's not just a Bayer/Foveon thing. It's a question of compromise
> between the evil of losing detail and the evil of aliasing
> artefacts. Different camera makers all with Bayer sensors adopt
> different compromises.

that's it exactly. you don't want an anti-alias filter so strong that
it eliminates all artifacts because it will eliminate details that
could otherwise have been captured. on the other hand, you don't want
to completely eliminate it because the images will suffer from
artifacts. the answer is somewhere in the middle.

foveon can get away with a weaker anti-alias filter because the
artifacts aren't anywhere near as ugly as with bayer, but sigma decided
to remove it entirely. in many cases, it may not matter whether there
are visible strands of hair or 5 lines versus 9 lines, but aliasing can
also be ugly low frequency beat patterns, and then the image is ruined.

External Since: Jun 03, 2004 Posts: 1804

"Chris Malcolm" <cam DeleteThis @holyrood.ed.ac.uk> wrote:
>
> Let's get away from the special case of regular stripes and consider
> wisps of hair in a headshot. Let's suppose there are nine wisps, and
> the anti-aliast picks up none of them whereas the aliast high-res
> camera picks up seven of them, getting their position and shape right,
> but of course misrepresenting their thickness in a "random" fashion.
>
> Which was more accurate?

In real images, hair and fur are one of the places where inadequately
antialiased images look really horrible because the aliasing artifacts
(jaggies, varying apparent thickness). So people really do mind that what
appears is different from what was actually there.

That reminds me: power lines in inadequately antialiased images are a joke
too: single strands turn into twisted pairs that twist at the characteristic
frequency resulting from the angle with which the power line crosses the
sensor matrix.

Bascially, you really don't want to look closely at fine fairly straight
lines imaged without adequate antialiasing.

David J. Littleboy
Tokyo, Japan