Image processing and computer vision today are faced with increasing big data applications. Excessive collection of Image data sometimes can have bad quality due to errors at the time of acquisition or at the time of transmission,  so for that problem the method is needed to perform image restoration. Image restoration is a process to make improvements to the image with the aim of obtaining a clean image from noise like the original image. Among the methods that can be used in image restoration, Markov Random Field (MRF) based on a probabilistic representation of image processing problems, namely maximizing the probability size calculated starting from the input data for all candidate solutions can provide a faster sub-optimal solution for image restoration.  Based on the implementation this experiment conducted with the noisy test image, the MRF method was capable to improve the noisy image up to 96.75 percent close to the original image without noise