The author is caching images to offer their users the privilege of not having to hit a Google server - and have to trust a Google server.
The stealing isn't what you think - or even what the author thinks: they didn't steal the images from Google. The images never belonged to Google to begin with, and Google was serving them up for free.
What the author did was deny Google an opportunity to track the users hitting Google servers. In other words, the author stole a bit of datamining revenue from Google.
Good on them!