In this paper, a new motion compensation approach for inverse synthetic aperture radar (ISAR) is proposed based on time-frequency analysis and image processing to the time-frequency image. In the approach, the velocity estimation is obtained by traditional curve-fitting technique, but the acceleration estimation is obtained by image-processing technique and curve-fitting technique together. Image-processing technique is used to get the time-frequency distribution of a chosen range cell after initial velocity compensation, and curve-fitting technique is used again to get the slope, which is proportional to acceleration. Computer simulation shows the accuracy of the estimated motion parameters by the new approach is satisfied for ISAR imaging, although the simplest time-frequency analysis method, i.e. short time frequency transform (STFT) is used. At the same time, the method is very competitive in speed with others for real-time processing of ISAR data.