GD can do this (and more). Here's a quick 'n' dirty example that will create a 100x100 pixel PNG and spew it to STDOUT (so make sure you redirect to a file in your shell).

use strict;
use GD;

my $im = new GD::Image(100,100);
my @colors; # Preallocate colors, since PNG has a limited number of co
+lors
for (0..255) {
  $colors[$_] = $im->colorAllocate(int(rand(256)),int(rand(256)),int(r
+and(256)));
}

for my $x (0..99) {
    for my $y (0..99) {
        $im->setPixel($x,$y,$colors[int(rand(255))]);
    }
}
binmode STDOUT;
print $im->png;
[download]

Before your print add:

print "Content-Type: image/png\n\n";
[download]

and your script is now a CGI script that can be put behind a webserver. There is now no need for a temporary file!

Thanks, myocom for the quick intro to the awesome GD module! It is a very useful module for me, and the documentation is very lengthy, however very useful.

Thanks tilly for showing me how to make this useful for CGI, since I tend to program in CGI more often than the shell usually.

I will be continuing to experiment with the GD module, and see if I can't get really good at utilizing this module for some really cool Perl artwork *hopefully*!

Andy Summers

I was actually sitting on some code for this for a while, and your post was finally the inspiration for me to get to work and add it to Data::Random. I just uploaded the new version (0.03), so it might be a while before it hits all the mirrors. When you get it, take a look at the rand_image() function.

Let me know if you have any feedback.

If instead you are looking for a way to use X from your windows machine, there are a number of good applications. One of the commercial apps that I like is Exceed by Hummingbird. A nice free app is VNC, which is a little like PCAnywhere, but works on more than just Windows.

elbieelbieelbie

When I read this, I immediately thought of this node.

Sometimes, this neural network that rides atop my shoulders is even better than SuperSearch...

There certainly is... all you need to do is put a link to the "image" into your html, e.g. <img src="random.pl" height="100" width="100" />.

What I did on one of my servers is actually to tell Apache that all the *.png-files in a certain directory are to be handled as perl-scripts, so the user on the other end thinks s/he's getting real *.png images.

Both methods in this post require your script to print "content-type: image/png\n\n"; before the actual image.

Hope that helps
BrotherAde

Since other people have talked about the mechanics of writing PNGs with perl, I figured I'd talk about the random-pixels bit (since that's fairly close to my area of specialization).

If all you want is random, unrelated pixels, that's easy enough to do, but you'll probably want some way of restricting their intensity and color (to make it easy to generate a bunch of pixels that are all mostly red and fairly dark, for instance -- in my experience, once you have a random image generator with a bunch of knobs and widgets, you spend hours playing with it :-). You might also want to be able to set "gain" on the data -- higher gain would mean more pixels at high and low intensities, lower gain would mean most of the pixels end up at moderate intensities.

A more interesting (to me) kind of random image data is Perlin (or coherent) noise. Well-implemented Perlin noise has a bunch of cool statistical properties, but at a glance it's just smooth noise. I'd rant about all the cool things you can do with it, but I have deadlines to attend to, so I'll leave you with a url: this page, by Hugo Elias.

If you're really interested in this stuff, the canonical reference is Texturing and Modeling: A Procedural Approach, which is damn good.