Blurring coherence events is the result of applying many spatial and temporal filtering algorithmswhen they are applied in order to suppress background random noise. Bayesian Filtering (BF) also suffers from mentioned problem. This paper develops a method for optimizing BF by Adaptive Neuro-Fuzzy Inference System (ANFIS) and Fuzzy C-Mean (FCM)clustering. First the structure of the GPR image is extracted using FCM. The structure and output of the BF for a random part of the data are used to produce output values for training ANFIS and after that, by generalizing the trained network to all data, filtered data would be achieved. The proposed method is applied on synthetic data-sets as well as tworeal 2-D GPR images gathered in an environmental study project. Performance of the method is evaluated by comparing the results of the proposed method to the output of BF. In synthetic data, the SNR value improved 63 percent more than of BF’s outputandthe visual comparison of the results are suggesting better performance in noise cancellation and resolution enhancement, both in synthetic and real data-sets.