This paper is concerned with the efficient compression and feedback of channel state information (CSI) in downlink multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems. Inspired by video coding, we propose a novel CSI compression and feedback scheme, referred to as hybrid transform coding (HTC). HTC consists of two coding types, i.e., selective time-domain coding (STDC) and differential time-domain coding (DTDC), which are adopted to exploit the correlation of CSI in both the frequency and time domains. We first develop closed-form expressions for the overhead-distortion performance of these two coding types in HTC. The parameters involved in HTC are then optimized based on the analytical results. Finally, the system-level performance of HTC is evaluated in both maximum eigenmode beamforming (MEB)-based single-user MIMO (SU-MIMO) and zero-forcing beamforming (ZFBF)-based multiuser MIMO (MU-MIMO) under Long Term Evolution Advanced (LTE-A) Release 10-based cellular networks. Simulation results show that HTC can significantly outperform the available alternative.