The development of a state estimator for a crystallization process, equipped with a sensor for the crystal size distribution (CSD), is discussed. The estimator is designed on the basis of a nonlinear distributed parameter model, which describes both the dynamics of the crystal size distribution and the output of the sensor. The process model is lumped to a set of nonlinear first-order differential equations, which are linearized leading to a high-order state-space model. With this model, a constant error feedback gain is designed by using Kalman filter theory. Estimates are used for the measurement and process noise covariance matrices. The designed gain is implemented in the nonlinear lumped process model. The resulting state estimator is evaluated with raw data from a pilot crystallizer, equipped with an on-line CSD sensor. For different sets of output data, the designed estimator is able to track the process output signal trend, while reconstructing the CSD, the supersaturation level, and a set of related variables. Sufficient robustness is demonstrated for sensor failure, initial state errors, and process disturbances.