Inverse synthetic aperture radar (ISAR) images generated of targets which have a large angle of rotation or non-linear rotational motion are blurred if the rotational motion is not taken into account. If the rotational motion is known, the data can be corrected by resampling from the polar grid on which the data is collected, to a rectangular grid. A focused ISAR image can be then formed by using a 2D FFT. In practice the rotational motion is not known a priori and so it must be estimated from the data itself. This paper presents a technique for estimating the target's rotational motion. The rotational motion is modelled as a quadratic, the parameters of which are set by minimising the entropy of a single cross range profile of the data. The minimum entropy cross range profile is the best focused cross range profile which should result in a focused ISAR image. Alternatively the entropy of the image itself could be minimised, but this has not been considered as it is computationally intensive. >