Please see Parsing HTML/XML with Regular Expressions for why it is indeed not a good idea to do this without a proper parser, especially look at the "spoiler" for lots of cases of perfectly valid HTML that will not be fun to parse with a regex. Here's an example with Mojo::DOM:
use warnings;
use strict;
use Mojo::DOM;
my $html = <<'ENDHTML';
<html><head><title>Title</title></head>
<body>
<div><div>
</div></div><div><div class="blue"></div></div>
</body>
</html>
ENDHTML
my $dom = Mojo::DOM->new($html);
$dom->find('div > div.blue')
->each(sub{ $_->parent->remove });
print $dom;
__END__
<html><head><title>Title</title></head>
<body>
<div><div>
</div></div>
</body>
</html>
I had a quick look at "Git for Windows", and it happens to include HTML::Parser. In the above thread, tangent showed an example with that module here, and because it's a fairly old but good module you will find lots of examples with it online as well. That Git distribution also appears to contain cpan as well, so you could try installing Mojo::DOM.