Image segmentation is a classical problem in computer vision and has been widely used in many fields. Due to the uncertainty in images, it is difficult to obtain a precise segmentation result. To deal with the problem of the uncertainty encountered in the image segmentation, an evidential Markov random field(EMRF) model is designed, which combines the evidence theory, a powerful tool for modeling and reasoning uncertainty, with the Markov random field. Based on EMRF, a novel image segmentation algorithm is proposed. EMRF uses evidential label field to describe the ambiguity of labels and distance of evidence to describe the relationship between labels of the neighboring pixels. The iterated conditional modes(ICM) algorithm is used for optimization. Experimental results show that the proposed algorithm can provide a better segmentation result against traditional MRF, fuzzy MRF(FMRF) and traditional evidential approaches.