The Otsu’s method is an effective algorithm for image segmentation. But it is unfit for the real-time image processing in the high frame rate infrared imaging tracking system for it is time-consuming. A fast half-approximation algorithm of Otsu’s method was presented based on the curve of the between-class variance. The optimal threshold was obtained from the first derivative of the between-class variance function with the half-approximation method. The theoretical analysis and simulation results show that the algorithm reduces the computation cost greatly. It has a good performance on real-time processing and is easy to be realized in engineering.