A Modified Metropolis Dynamics (MMD) simulated annealing algorithm based on Markov Random Field (MRF) is proposed. Threshold α of the proposed algorithm optimized Metropolis arithmetic. Based on the Beyes theory, the segmentation issue is transformed to the Maximum A Posteriori (MAP) and the parameter prediction algorithm is provided. In the experiment, the proposed algorithm is compared with the traditional Metropolis algorithm, the Gibbs sampler and Iterated Conditional Mode (ICM), and the results show that the segmentation efficiency and segmentation accuracy of the proposed algorithm improve obviously. Through concrete CT image segmentation experiment, it is also found that the inner structure and defect of the concrete materials can be integrally reflected based on MMD.