The point is to spare output pins/drivers.

So you connect the LEDs into a matrix. Then you light up the needed LEDs only in one row at a time - in step one light up required LEDs only in first row, in step two light up required LEDs only in second row etc. - and repeat this in a fast rate, one full cycle (a "frame") within less than say 1/50 second. The persistence of vision makes your brain believe, that all of the rows are lighted up at once.

The moving sign is the same as TV - you simply change the pattern of light-up LEDs in each frame.

This is just the basic principle; it has many many pitfalls and problems - have a look e.g. at Payam's LED sign project - it has been discussed in hundreds of posts already here.

Jan Waclawek