A new technique for the on-line estimation of spot volatility for
high-frequency data is developed. The algorithm works directly on the
transaction data and updates the volatility estimate immediately after the
occurrence of a new transaction. We make a clear distinction between volatility
per time unit and volatility per transaction and provide estimators for both. A
new nonlinear market microstructure noise model is proposed that reproduces the
major stylized facts of high-frequency data.