An algorithm combining the information entropy and region growing method is put forward for the accurate segmentation of the IR small target image in background of fluctuation.First,the IR original image is filtered using the adaptive smooth filter after analyzing the image features,thus the Gaussian noise is restrained without destroying the strong edge of the target.Then,the image segmentation is implemented using the maximum entropy algorithm based on the Gray level-Gray absolute difference (G-G) histogram,by using the feature that the target edge is high in gray level and changes rapidly in its surrounding region.After the image segmentation,the seed points are selected according to the threshold obtained in the maximum entropy algorithm.And the region growing method is realized in the segmented images.Experiment results show the effectiveness of the methods given here in target-detection under background of fluctuation.