Traditional methods can extract target contour features from inverse synthetic aperture radar （ISAR） images. However， they have the problem of low accuracy. Therefore， this article proposes a target contour extraction method from ISAR images based on an improved Clean algorithm. Based on the random occurrence of noise， this article establishes an optimization function for scattering point extraction， and uses the iterative Clean method to extract as many weak scattering points as possible. A scattering point quality evaluation method based on connectivity analysis is introduced. This method is used in the post-processing of the Clean algorithm. It denoises the points extracted by the Clean algorithm and preserves the scattering points of the target. The scattering point set is processed with alpha-shape and wavelet filtering to get the target contour. The simulation and measured data processing results show that this method gives more accurate contour than existing methods. It also has better noise resistance.