To effectively segment the targets of interest with blurred boundaries and low contrast in infrared images, an infrared image segmentation model based on region competition is proposed. For the Geodesic Active Contour (GAC) model sensitive to noise, assuming that the object and background in an image obey Gaussian distribution. Then, according to the probability of pixels belonging to the infrared object, a region energy term is constructed so as to improve the robustness of the model. Finally, a signed distance function is introduced to avoid the re-initialization of the curve in the process of evolution and improve the efficiency of the model. The experimental results show that the proposed method can segment the targets of interest in an infrared image effectively.