A new adaptive time-frequency Inverse Synthetic Aperture Radar (ISAR) image analytical method is presented, which combines the ISAR processing with the time-frequency analysis algorithm-Adaptive Guassian Representation(AGR). This technique uses Guassian basis function to adaptively parameterize the data. And as a consequence, the point-scattering and dispersive-scattering mechanisms can be readily separated based on the width of the Guassian bases. Applying this transformation to range and crossrange axis of ISAR image can extract the frequency-dependent and the aspect-dependent features of nonpoint-scattering centers respectively. The method is tested by using simulation data and real data. The results show that it is a robust tool to post-process ISAR image. And it can provide additional information about the pointlike and dispersive scatting mechanisms that can be useful for ATR.