Heuristic for determining what is likely to be a logo on a website?

I'm building an algorithm for determining logos of websites, given nothing but the root url of the website.

What I'd like 开发者_C百科to do is to provide a collection of likely logos to my client, and allow them to choose the logo of the site.

I'm looking for the best techniques for displaying a logo on a webpage (most common sizes, most common image replacement techniques, etc), so that I can better tailor my algorithm to produce less images (but more likely logos).

Thanks in advance!

---

This won't be easy, but here are some things you can look for:

1. Look for an element with the word 'logo', or the name of the website (on www.google.com search for the word 'google') in the ID, Class or Filename
2. If several are found give priority to the first one you find (start from the top)
3. If nothing was found, search for h1 elements with background-image properties
4. Search for any elements with the word 'logo' as an ID and check for background-image property