A digital-signal-processing technique is presented that can compensate for the frequency-independent inphase/quadrature (I/Q) mismatch and dc offset in zero-intermediate-frequency direct-conversion receivers. The proposed compensator consists of dc offset and I/Q mismatch estimators followed by a dc offset canceller and self-image suppressor. The estimators are nondata-aided (NDA) schemes that are derived by assuming that the desired signal is a random variable having an arbitrary circular symmetric distribution. The Cramer-Rao lower bounds (CRLBs) for NDA estimation are derived, and the characteristics of the estimators are analyzed in terms of mean-square errors. In particular, it is shown that for white Gaussian distributed signals, the accuracy of the estimates approaches the CRLBs as the number of observed samples increases. The characteristics of the proposed method as well as the validity of the analytical results are demonstrated through computer simulation.