Implementation of robotic systems has significantly improved the flexibility and accuracy of prostate brachytherapy. In our previous study, an ultrasound (US)-guided dual-arm robotic brachytherapy system was developed. This system was integrated with an end-effector for needle insertion and an end-effector for US probe. The calibration accuracy of the system determines the effectiveness of the whole system. However, existing calibration methods are mechanism-based calibration methods or image-based calibration methods. With these methods, external tracker and complex phantom are needed, which may result in accumulative errors. Therefore, this paper presents a mechanism-image fusion approach to calibration of a US-guided dual-arm robotic brachytherapy system. With this approach, no tracker or complex phantom is needed as the dual-arm robotic system can use one arm as a substitute for tracker and phantom to calibrate the other arm. The needle-tip positions localized by the mechanism and image are utilized while the needle is also the controlled object of needle insertion. Experiments using three registration algorithms were performed, and results were evaluated utilizing the leave-one-out cross-validation method. The results showed that the calibration accuracy of the whole system is 0.650.31 mm. Additional experimental and parametrical comparisons of the proposed method with the state-of-the-art methods were conducted, and the results justified the outperformance of the proposed method.