The intensity inhomogeneity phenomenon often appears in actual images, making their segmentation challenging. To efficiently segment such images, a level-set (LS) method with a multiplicative–additive constraint model is proposed. First, based on the observed image intensity property and the multiplicative–additive model, we define the dual bias fields, which explain the full nature of intensity inhomogeneities. Considering the ill condition of the model, herein, we further impose reasonable constraints while ensuring smooth change of dual bias fields within a predefined range. Additionally, we employ an increasing-sequence strategy in the numerical method to ensure balance between the constraint effect and LS evolution, further improving the robustness of our model. When compared with the traditional single bias–field models, our model presents more competitive capabilities. The experimental results on various images validate that compared with the state-of-the-art LS models, our model exhibits better performance in terms of accuracy and robustness.