Would using the css diff technique here make diff images easier to understand:

http://franklinta.com/2014/11/30/image-diffing-using-css/

(feel free to close this if it's not. I just wanted to point this out ot you. Also, they link to chrome extension's captureVisibleTab method and I wondered if it could be used to make a CSSCritic chrome extension that passes the rendered page back to the regression runner. NB: it's not currently supported by Firefox, and there's no bug in bugzilla for it yet, but the WebExtensions page says at the bottom that they plan to add "Ability to take images of frames/tabs (like canvas.drawWindow)" so I expect there will be a way to write a firefox extension to do it in time.)