Due to the absence of atmospheric interference, space-based terahertz (THz) inverse synthetic aperture radar (ISAR) can obtain high-resolution images of space satellites at long ranges, offering significant application potential. However, the intricate relative dynamics between the sensing platform and the target induce non-uniform variations in the radar-target aspect angle, resulting in severe image defocusing. This paper proposes a non-parametric THz-ISAR imaging algorithm for non-uniformly rotating space targets based on ambiguity function (AF) domain filtering. Firstly, the azimuth signal is transformed into the AF domain after translational motion compensation and range migration correction. A filter is then designed by analyzing the cross-term distribution within the AF domain. After eliminating the influence of cross-terms, time-frequency resolution is enhanced using time-frequency rearrangement techniques, ultimately yielding high-quality imaging results. The effectiveness of the proposed method is verified by simulation and experimental results, and the superiority of the proposed method is verified by comparison with the traditional algorithm.