This paper addresses the problem of channel estimation in multi-cell interference-limited cellular networks with multiple antennas, focusing on both finite and large-scale antenna regimes (massive MIMO). The paper presents a novel approach to mitigate the pilot contamination effect, which is a major bottleneck in channel estimation performance. The proposed method enables low-rate coordination between cells during the channel estimation phase, leveraging second-order statistical information about user channels to discriminate between interfering users. It is shown that in the large-number-of-antennas regime, the pilot contamination effect can be completely eliminated under certain conditions on the channel covariance. Simulations confirm performance gains over conventional channel estimation frameworks, even for small antenna array sizes. The paper introduces a Bayesian channel estimation method that exploits covariance information in the presence of pilot contamination. It demonstrates that the channel estimation performance depends on the overlap of dominant signal subspaces in the desired and interference channel covariance matrices. The method uses a covariance-aware pilot assignment strategy during the channel estimation phase, which helps shape the covariance matrices to satisfy the needed conditions for interference-free channel estimation. The paper also shows that the proposed method achieves performance close to interference-free scenarios, even with moderate numbers of antennas and users. The paper analyzes the performance of the proposed method in the large antenna number regime, showing that the pilot contamination effect can be eliminated when the signal subspaces of the desired and interference channels are non-overlapping. The analysis is based on the assumption of an uniform linear array with supercritical antenna spacing. The paper also discusses the impact of standard deviation of Gaussian AOAs on the estimation and shows that the estimation error increases with the standard deviation. The paper concludes that the proposed method provides significant performance gains in massive MIMO systems, particularly in the presence of pilot contamination.This paper addresses the problem of channel estimation in multi-cell interference-limited cellular networks with multiple antennas, focusing on both finite and large-scale antenna regimes (massive MIMO). The paper presents a novel approach to mitigate the pilot contamination effect, which is a major bottleneck in channel estimation performance. The proposed method enables low-rate coordination between cells during the channel estimation phase, leveraging second-order statistical information about user channels to discriminate between interfering users. It is shown that in the large-number-of-antennas regime, the pilot contamination effect can be completely eliminated under certain conditions on the channel covariance. Simulations confirm performance gains over conventional channel estimation frameworks, even for small antenna array sizes. The paper introduces a Bayesian channel estimation method that exploits covariance information in the presence of pilot contamination. It demonstrates that the channel estimation performance depends on the overlap of dominant signal subspaces in the desired and interference channel covariance matrices. The method uses a covariance-aware pilot assignment strategy during the channel estimation phase, which helps shape the covariance matrices to satisfy the needed conditions for interference-free channel estimation. The paper also shows that the proposed method achieves performance close to interference-free scenarios, even with moderate numbers of antennas and users. The paper analyzes the performance of the proposed method in the large antenna number regime, showing that the pilot contamination effect can be eliminated when the signal subspaces of the desired and interference channels are non-overlapping. The analysis is based on the assumption of an uniform linear array with supercritical antenna spacing. The paper also discusses the impact of standard deviation of Gaussian AOAs on the estimation and shows that the estimation error increases with the standard deviation. The paper concludes that the proposed method provides significant performance gains in massive MIMO systems, particularly in the presence of pilot contamination.