In this paper, to improve image performance of biomedical data, Markov Random Field (MRF) and Cellular Neural Network (CNN) structures are combined and a new approach, Markov Random Field-Cellular Neural Networks (MRF-CNN) is introduced. MRF-CNN structure can be applied to biomedical data for various image processing problems such as noise filtering, edge detecting, blank filing etc., with noise variance up to 9 dB and better results are obtained according to MRF and CNN schemes. In training of MRF-CNN, Recursive Perceptron Learning Algorithms (RPLA) is studied.