To comply with system requirements (e.g., power consumption, bandwidth) in many wireless sensor networks, the signals are roughly quantized. In this paper we address the case in which different sensors in the network sense autonomously a physical parameter field and are quantized in different quantization resolution. We present the Maximum Likelihood estimator (MLE) for general parameter estimation in such a case, and we study in details the special case of linear parameter estimation, for which we compare the MLE to the Naive MLE (NMLE) which disregards the quantization. While being asymptotically optimal, the MLE is a complex processor. Therefore, we suggest a sub-optimal estimator, simpler than the MLE, whose performance is close to the optimal one, being better than that of the NMLE. Simulation results demonstrate the operation of the different estimators in various conditions.