Rumour is that the next codec after H.264 will be wavelet-based. There's already an experimental one written by an individual (check the doom9.org forums).
The problem with making a wavelet video codec is that it is a lot less natural to implement things like motion compensation as there are no macroblocks to speak. Because of this, wavelet codecs generally have a lower SNR than DCT-based codecs at similar bitrates using current technology.
The difference is that DCT codec artifacts often show up as blockiness at the division between macroblocks, whereas wavelet-based codecs "spread out the error" more evenly. Some say that wavelet subjectively looks better than DCT-codecs despite having a lower SNR in absolute terms.
In any case, we first have to see how H.264 turns out. It's supposedly much better than MPEG-4 Part 2 (commonly referred to as only MPEG-4). Last time I checked the code that existed for H.264 was largely proof-of-concept code and required a ridiculous amount of computing powered for both encoding and decoding, so we'll see how it matures.