Abstract: By using the Kalman filtering method, a unified and general white noise estimation theory is presented for the first time. It can handle the filtering, smoothing and prediction problems in a unified framework for both the input white noise and measurement white noise m linear discrete time-varying and time invariant stochastic systems. The optimal and steady-state white noise estimators are presented, and white noise renovation filters and Winner filters are also presented. They can be applied to seismic data processing m oil exploration, and provide a new tool to solve the state and signal estimation problems. Two simulation examples show their effectiveness.