The use of multiple antennas at both the transmitter and receiver is well motivated by the substantial performance gains that can be realized, both in terms of an increase in achievable data rate and a decrease in error rate. The achievable data rate is characterized by the mutual information of the channel. In case of frequency selective multiple input multiple output (MIMO) channels, expressions for mutual information are obtained by using a block transmission model. In this paper, we consider a filterbank equalization framework, which is an attractive alternative to the block processing approach. We derive expressions for mutual information with the minimum meansquared error (MMSE) filterbank receiver, for both joint detection and symbol-wise detection. It is shown that the filterbank approach provides much better information rates than the block processing approach. Simulation results are provided to validate our analysis. © 2011 IEEE.