I'm also not a guru on how all this blockchain technology works under the hood. So if images were persisted on the blockchain, I wonder how that would affect witness nodes? That is, would each node have to store the entirety of the blockchain contents? If so, then how many tera/peta bytes would it take to store all the associated images from all the Steemit posts for all time? And then you also have to contend with video!
With the current way the blockchain is programmed, it would have to store everything - which would not be scalable if it had to store images, videos, etc.
They are looking into alternate ways to support it in the future though.