MPEG coding divides the screen into 8x8 pixel blocks, further grouped into 16x16 macroblocks. Then it encodes each block by essentially doing a Fourier transform to find how intensities change across the block. The 8x8 frequency coefficients are then divided by a frequency-dependent function to reduce the number of bits used; the whole lot can also get scaled down if the encoder wants to use less bits for compression control. There's also encoding of deltas from frame to frame, both in spatial terms, and in time within a block.

The upshot of this is that if there's a lot of change going on, the higher frequencies within a block have a tendency to get discarded, and you get a block that loses detail, ultimately tending toward a uniform colour.