For better using of the inverse synthetic aperture radar (ISAR) images, it is more desirable to estimate the rotational velocity (RV) for image cross-range scaling. To solve this problem, an algorithm based on affine projection matrix is proposed. This method consists of two parts. The first part is to transform the two sequenced sub-aperture images by affine matrix. The aim is to eliminate the negative effects caused by ‘scaled–stretched’ transformation between the two images. In the second part, the processed images are mapped into polar coordinates system. In polar coordinates, the rotational motion in ISAR image is converted into the translation in the angle dimension. The correlation coefficient is used to search the right RV which provides the maximum correlation between the two processed images. Finally, experimental results based on simulated and real ISAR data are used to prove the effectiveness and practicability of the proposed method.