In full-duplex orthogonal frequency-division multiplexing (OFDM) systems with in-phase and quadrature (IQ) imbalances,a time-domain least squares (TD-LS) channel estimator is proposed for estimating both the source-to-destination (intended) and the destination-to-destination (self-interference) channels.To further improve the performance,an adaptive orthogonal matching pursuit (OMP) is proposed and its sparsity is estimated by a threshold method.Finally,the full-duplex interference is removed using serial interference cancellation and sphere decoding is used to complete the maximum likelihood detection.Simulation results demonstrate that in terms of both the mean square error (MSE) and the bit error rate (BER),the proposed adaptive OMP performs better than the TD-LS by exploiting the sparse property of the channel.Additionally,compared to the gradient projection,the proposed adaptive OMP is better in the low and medium signal-to-noise ratio (SNR) regions and marginally worse in the high SNR region.