Axiom Home Page Forums General Discussion Technical Questions Itunes, bitrate, and other questions...

Fitting in a square; that's actually pretty insightful.

I'll give.

The Fourier transform can only be applied to 1 dimensional data. The change in amplitude of audio samples over time is what is fed into it for MP3s. But pictures are 2 dimensional. So what JPEG did was to break each picture into squares, 8x8 being the most common. Then they needed a standardized way to linearize the pixels in that block. Keep in mind the Fourier transform works best when the next sample fed into it is closely related to the previous. So what's the path through an 8x8 grid which most closely relates each next sample to the previous.

There's your answer!