A time-domain（TD） least square（LS） channel estimator is first proposed to estimate channel parameters of OFDM system with IQ imbalances at both transmitter and receiver.Then,an iterative shrinkage（IS） algorithm from compressed sensing is adopted to further improve the estimation performance by using the TD-LS solution as the initial value of IS in the case of sparse channel.Simulation shows that our algorithm combining TD-LS and IS performs better on bit error rate than the frequency-domain LS and matching pursuit in sparse Hilly Terrain channel when the same LS equalizer is adopted.