Observed high-frequency prices are contaminated with liquidity costs or market microstructure noise. Using such data, we derive a new asset return variance estimator inspired by the market microstructure literature to explicitly model the noise and remove it from observed returns before estimating their variance. The returns adjusted for the estimated liquidity costs are either totally or partially free from noise. If the liquidity costs are fully removed, the sum of squared high-frequency returns - which would be inconsistent for return variance when based on observed returns - becomes a consistent variance estimator when based on adjusted returns. This novel estimator achieves the maximum possible rate of convergence. However, if the liquidity costs are only partially removed, the residual noise is smaller and closer to an exogenous white noise than the original noise. Therefore, any volatility estimator that is robust to noise relies on weaker noise assumptions if it is based on adjusted returns than if it is based on observed returns.