They come from legal documents submitted to the courts, so they were probably greyscale PDFs that had been printed, scanned and printed. If someone could have screenshotted the PDF and included it in a Word document I'm sure they would have somewhere during their lifespan.

Documenting things for inclusion in litigation isn't cheap. There's no reason they needed to print the images on glossy photo paper in full color. Google is asking Oracle to pony up over $4 million just for their bill of costs over the Java API Copyrightability case: http://www.groklaw.net/articlebasic.php?story=20120706111715...