When sharing a link in major websites like Digg and Facebook; it will create thumbnails by capturing main images of the page. How they catch images from a webpage? Does it included loading the whole page (e.g. by cURL) and parsing it (e.g. with preg_match) ? To me, this method is slow and unreliable. Does they have a more practical method?
P.S. I think there should be a practical method for quick crawling the page by skipping some parts (e.g. CSS and JS) to reach src attributes. Any idea?
They typcailly look for an image on the page, and scale it down on their servers. Reddit's scraper code shows a good deal of what they do. The Scraper class should give you some good ideas on how to tackle this.
JohnD's answer shows that Reddit uses embed.ly as part of their Python solution. Really embed.ly does the hard part of finding the image and they're free under 10,000 requests/mo.
They generally use a tool like webkit2png.
Some use
<link rel="image_src" href="yourimage.jpg" />
included in the head of the page. See http://www.labnol.org/internet/design/set-thumbnail-images-for-web-pages/6482/
Facebook uses
<meta property="og:image" content="thumbnail_image" />
来源：https://stackoverflow.com/questions/7462044/how-major-websites-capture-thumbnails-from-a-link