Resizing gifs?

[Geoff Hart <ghart -at- videotron -dot- ca>]

Elizabeth O'Shea wrote: <<I'm sure this is simple, but...>>

Nothing about digital images is simple. Software should solve your 
problem for you--and indeed, tools such as Smooth Scaling from Wextech 
and RoboHelp's tool for smoothly resizing graphics do help--but you 
still have to jump through hoops sometimes.

<<We have a gif that's 1668 x 545 pixels. We also have the psd of this 
file. I've been asked to resize it to about 330 x 108 pixels. The 
resultilng quality is horrible.>>

First off, make sure you keep a safe backup copy of the original 
Photoshop (PSD) document. Wherever possible in image editing, work from 
the original image, not from a transformed image (such as the GIF 
you're trying to resize). Each time you transform a graphic, you lose 
data that you often can't get back.

<<I have: <snip several options>... changed the image size of the pds 
in Photoshop an used the SAve for Web option to save it as a gif - 
opened the gif in Illustrator, saved it as ai, resized it, and saved it 
for web. They all produce horrible quality.>>

The problem is twofold. First, there's a mismatch in resolution: do the 
math (original divided by final dimensions) and you'll see that any 
reduction near the value you want doesn't produce an integer value for 
both dimensions. This means that the software is trying to calculate 
the values for a fractional pixel and failing. You'll have better luck 
if you choose a final resolution closer to an integer value (e.g., 
1668/4 = 417 pixels, which is close to what you want, but then you 
don't get an integer value for 545/4).

The second problem is one of scale: details that are sharp at 100% 
magnification are likely to be invisible at 20 or 25% magnification, 
particularly if those reduced details become smaller than 1 pixel in 
any dimension. So take a close look at your original image: any 
features that are smaller than about 5 pixels will disappear when you 
reduce the image (because they'll be too small to display in a single 
pixel). Can you simplify the original image to eliminate all such 
features? Can you enlarge all such features so that they're at least 5 
pixels in each dimension? In short, simplify the image so that it'll 
work at the lower size.

Bottom line is that you're likely to get a fuzzy image from the current 
bitmap no matter what degree of resizing you try, since the x and y 
dimensions aren't easily divisible by the same number and since some 
features will vanish or blur because they're too small to remain 
visible at the smaller size. Probably the simplest possibility would be 
to crop the image to 1665 pixels so that both dimensions can be resized 
identically by a factor of 5 (to 333x109)--the lost of three pixels in 
the original should be undetectable at the new resolution. However, 
this won't solve the problem of features that shrink below 1 pixel in 
size.

Alternatively, redraw the figure from scratch so that each element is 
at the target size, or at some nice multiple thereof so that it will be 
at least 1 pixel high in the final image; if the image contains 
anything like vector art (e.g., rectangular or circular elements as 
opposed to random dots and patterns--bitmaps!), use a vector program 
such as Illustrator instead. That lets you scale the image far more 
smoothly to whatever size you want before converting it to a bitmap.

--Geoff Hart   ghart -at- videotron -dot- ca
(try geoffhart -at- mac -dot- com if you don't get a reply)