Multisampling is a feature that is designed to reduce noise
by averaging the pixel values from multiple samples. The theory
is that since noise is inherently random, noise will cancel
itself out when the values are averaged over multiple readings.
In this section, we will examine a slide to study the effects
of using multisampling on images that aren't manipulated as
well as on images that are heavily manipulated.
The methodology used involves scanning the same slide twice
in 14 bit mode and Adobe RGB color space on a Nikon Super
Coolscan 4000. One scan is made without multisampling and
the other is made with multisampling set to 16X. These images
are then cropped at "actual pixels" size, converted
to sRGB color space, converted to 24 bit (8 bits per channel)
images, and saved out as high quality (Q = 90) jpgs in Photoshop
to make them viewable on the web. The 2 original scans then
have their histograms adjusted in Photoshop (in 16 bit mode)
to study the effects of Multisampling on images that are heavily
manipulated. Here, their white points are set to 200 and their
gammas are set to 1.4. The exporting process is repeated to
make these images viewable on the web. This system is not
perfect due to the compression of the images and other factors
but I believe it accurately shows the differences (or lack
thereof) between using Multisampling and not using Multisampling.