A method is proposed to reduce the requirement on the scanning speed for dynamic computed tomography of relatively fast moving objects such as the heart. From the projections measured at different times, a set of optimal projections is computed for reconstructing the image at any specified time within a permissible interval. In the method, the corresponding characteristic features of the measured projections at the same viewing angle are matched optimally by a nonlinear transformation. The matched projections are then used in an interpolation procedure to compute the projections required at the specified time. The method has been tested with mathematical phantoms that undergo partial expansion, contraction, and translational motion. The results show that the proposed method can provide acceptable images from data collected at scanning speeds that would otherwise produce severely blurred images by conventional methods.