In radar processing, inverse synthetic aperture radar (ISAR) takes advantage of the target rotational motion to provide a 2-D image of the target. The “quality of this image must be sufficient to be used for target recognition. However, among the main problems to be addressed, the translation motion which is another component of the target motion must be foremost compensated. If this is not the case, a blurred image is obtained. In this paper, we propose a new approach for translation-motion compensation based on the target radial velocity estimation. This method consists in characterizing the radar echoes by several features (mean and wavelet coefficients for instance) and in storing them in sequences which represent the parameter evolution over time. Spectrum analysis methods are then used to get Doppler frequency estimates. To make the estimation robust against additive noise and sea clutter, several estimates are combined to deduce the target radial velocity estimation. Tests on experimental data confirm the validity of the approach.