The main task of infrared image dehazing algorithms is to solve the problems of low visibility and blurring in infrared images; these problems arise from Mie scattering. However, current infrared image dehazing algorithms poorly estimate the dark transmittance of infrared images. Hence, in this study, an infrared image dehazing algorithm is developed based on the dark primary color prior of the haze-line. First, the Hough transform was employed to estimate the atmospheric illumination. Second, a dark primary color prior was employed to address the failure of the haze-line dehazing method in some scenarios. The transmittance was estimated by assuming that the dark end of the haze-line corresponds to the real color, and a transmittance map was obtained. To remove noise in the transmittance map, total variation regularization was used; thus, the transmittance map was optimized. The experimental results obtained using LTIR, a public infrared dataset, as the test dataset show that the proposed algorithm can enhance the clarity of infrared images without affecting the distribution of infrared radiation; in addition, the results show that the proposed algorithm enhances infrared images corresponding to various scenes. The proposed method accurately estimates transmittance and effectively dehazes infrared images.