With a bit of thinking you can do most of the work with shift registers. With HC595's you can have the output stable (as in displaying the current data) whilst you are shifting the next set of data in during the time you have the leds turned on (say around 1mS) You divide the whole display into smaller sections and concurrently multiplex each of these sections. You could even have a micro per section and tie them together serially - and have one micro control a string of slave micros.