How to Optimize the Processing Efficiency of a Video Processor?

In many applications you often deal with vividas and ti highly corrlated data, which are expansive. But the good news is that there are a lot of redundances into data. So processing also, i believe, shouldn't grow lenear, i t the amount of data. How can you cret one building blocks of a euronetri processor, where you can a levera disinherit redundance in your data, somehow factorize itd and just don't compute the same thing over and over? Maybe sort of prevent computing similar things and reduce what has been processedd for any task general.