A group of MIT researchers believe they've found a way to speed up audio, video, and image compression by improving on the Fourier transform. If you're not familiar with the method, it provides a way for complex signals to be broken into their composite parts and then recompiled. It was improved on in the mid-60s with the Fast Fourier Transform (FFT) algorithm, which, as the name suggests, made the whole process much quicker and practical to use. FFT has been used in compression and other fields already — once you split up a transmission into its composite parts, you can forget the less important frequencies and just send along the major players, reducing file sizes. This works for compression because most media is "sparse." For example, the researchers found that in a block of 64 pixels from a larger image, 57 could be discarded on average with little to no effect on the overall picture. With the researchers' new findings, they say there's now an even faster method to split signals and determine what's negligible.
The researchers claim to have improved on the FFT by more efficiently using filters to isolate smaller individual slivers of the signal which typically only have one predominate frequency. Once they've singled out a bit of the signal, there are two different methods (explained in two different papers) for determining what can be discarded. One keeps cutting down that sliver of data and retains the parts with the most powerful signal, and the other borrows a technique used in 4G data networks, called OFDMA. The researchers claim that most kinds of data that you'd care about (including audio, images, and video) are "sparse," and therefore can be compressed with the new algorithm up to ten times faster. While there's no commercial application that uses the new algorithm yet, it is a promising development that may help us stay under our mobile data caps in the future.