From: Pieter Bos <bos.pieter_at_gmail.com>
Date: Fri, 17 Nov 2006 12:43:06 +0100

"Christopher Woods" <christof_at_infinitus.co.uk> wrote in message
news:007801c709fa$c51ffb70$4001a8c0_at_20hz...
> I guess a comparable analogy (best I can think of right now)
> would be if you take a large image, and resize it to a smaller size - if
> you
> have your settings to just go with each pixel's nearest neighbour when you
> shrink the image, you'll get uneven lines, jaggedy edges and it'll look a
> bit poor... Whereas if you set your image program to do bicubic or
> bilinear
> resizing, it looks at the pixels, their relationship to the ones next to
> them, and 'redraws' the image in a sense, blurring together areas for want
> of a better description, to produce a more aesthetically-pleasing result -
> lines look smoother, colours blend and gradient better...
>

This analogy is more like increasing the time-resolution (sample rate) than
increasing the bit-depth.

Perhaps a better analogy, speaking in terms of images, is exactly the same:
dithering. If you have only a limited number of colours, you can produce
more colours by drawing tiny dots of different colours. Our eyes will
automatically blend the colours, so we see a colour that is not actually
there but which is made up of a blend of the other colours. This is applied
for example in computer screens with a limited number of colours and in
printers - you can see the dots in every photo in every newspaper!

You can try this yourself: Get two different colour pencils and draw lots of
dots on a piece of paper, alternating between both the colours.Make some
different areas, some with more dots with one colour, others with more dots
with the other colour. Make sure to draw the dots close together. Now look
at it from a distance. You will see different colours! You can try making a
drawing with this by changing the number of dots.

If you have only one pencil available, just make drawings with dots, but
change the number of dots in the same area (eg. completely filled, half
filled...).

In audio, the concept works almost the same. Think of the bit-depth of the
audio (16 bit...) as the number of available colours. If you want to match a
value that is not there, you can add some rapidly changing values close to
the value you want to match to gain a slightly better quality audio and a
bigger dynamic range. Now, you're performing dithering!

Pieter
Received on 2006-11-17