" often they are not embeded in the website but just a link."
Not too sure what this sentence is supposed to mean... Images in a Web page are always referred by a link. Sometime, the link points to the site that also hosts the page. Sometime it points to another site, but it is rare, because unless the site is made for such usage, it is seen as bad practice (consuming somebody else's bandwidth).
But whatever the site hosting the image, the way to get the image is the same.

To find these images, you just need to parse the HTML code (there are Java libraries to ease the task) and to find the <img> tags (although technically there can be background images declared in CSS too). Finding if an image belongs to an article or is just site decoration is trickier, though.