A run-length encoder is generally going to do worse on two images than one, and it's already bad on one.
As the other answers have elucidated, there's a spectrum of fidelity (or bandwidth) to transmit the clandestine data, but there's also a obscurity/security spectrum which depends on your intended/perceived adversary.
As two opposite ends of this multidimensional spectrum: you could hyper-compress then encrypt your payload and sprinkle its bits one by one on the LSB of every N pixels (or even a varying number of pixels based on a key or hash of the previous bits), or you could do what Navin's great suggestion for a very simple system.
Because the carrier effective bandwidth is so ridiculously high, you could even take the 2 lowest bits of the 3rd pixel and make them a type, so you could pass text, images, files, and more,
I'd also like to point out that steganography can also be a great tool for good when used to bypass oppressive governments and regimes.