The high heterogeneity of petroleum reservoirs, represented by their spatially varying rock properties (porosity and permeability), greatly dictates the quantity of recoverable oil. In this work, the estimation of these rock properties, which is crucial for the future performance prediction of a reservoir, is carried out through a history matching technique using constrained ensemble Kalman filtering (EnKF) and particle filtering (PF) methods. The first part of the thesis addresses some of the main limitations of the conventional EnKF. The EnKF, formulated on the grounds of Monte Carlo sampling and the Kalman filter (KF), arrives at estimates of parameters based on statistical analysis and hence could potentially yield reservoir parameter estimates that are not geologically realistic and consistent. In order to overcome this limitation, hard and soft data constraints in the recursive EnKF estimation methodology are incorporated. Hard data refers to the actual values of the reservoir parameters at discrete locations obtained by core sampling and well logging. On the other hand, the soft data considered here is obtained from the variogram, which characterize the spatial correlation of the rock properties in a reservoir. In this algorithm, the correlation matrix obtained after the unconstrained EnKF update is transformed to honour the true correlation structure from the variogram by applying a scaling and projection method. This thesis also deals with the problem of spurious correlation induced by the Kalman gain computations in the EnKF update step, potentially leading to erroneous update of parameters. In order to solve this issue, a covariance localization-based EnKF coupled with geostatistics is implemented in reservoir history matching. These algorithms are implemented on two synthetic reservoir models and their efficacy in yielding estimates consistent with the geostatistics is observed. It is found that the computational time involved in the localization-based EnKF framework for reservoir history matching is considerably reduced owing to the reduction in the size of the parameter space in the EnKF update step. Also, the geostatistics-based covariance localization performs better in capturing the spatial heterogeneity and variability of the reservoir permeability than the traditional methods. In the second part of the thesis, we extend the history matching implementation using the particle filtering. Reservoir models, being nonlinear, the distributions of the noise and parameters are generally non-Gaussian in nature. Since the EnKF may fail to obtain accurate estimates when the distributions involved in the model are non-Gaussian, we attempt to use a completely Bayesian filter, the particle filter, to estimate reservoir parameters. In addition, the geostatistics-based covariance localization is also coupled with the particle filter and is found to perform better than the filter without any localization.