Autofocus algorithm is significant to acquire high quality Synthetic Aperture Radar (SAR) images. In order to focus the SAR images with complex scene and complex motion error, this paper proposed an autofocus method based on the minimum-entropy criterion and genetic algorithm. In the method, the phase error in the SAR image is modeled by the polynomial and the entropy of the images is the cost function. Then the cost function is minimized by the genetic algorithm to obtain the optimal estimation of the phase error. The proposed method is especially suitable for the problems that there are less isolated prominent points and the high-order phase error is not negligible. The simulated data and the real data results have validated the method.