In multispectral image mosaic, due to the imaging requirements and the information differences of different band images, it is impossible to complete the fast mosaic of full band spectral images under the condition of low overlap rate. To solve this problem, an adaptive low overlap multispectral image fast mosaic algorithm based on phase correlation enhancement is proposed in this paper. Taking the maximum information entropy channel of spectral image as the reference channel, an image correlation enhancement segmentation method is designed to cut off the specific step of two adjacent lens spectral images under the reference channel to improve the overlap rate of spectral images to be matched. The least square method is introduced to dynamically adjust the number of divisions and the division step length to realize the optimization of the relative offset while improving the realization and rationality of the algorithm. Using the principle of consistent offset of two adjacent sub lens images under different channels, through the positioning mapping of the displacement of the reference channel, the problem that individual channels cannot be spliced due to insufficient information is solved, and the splicing efficiency of multispectral images is improved. Experiments show that this method can realize the fast mosaic of 16 channel spectral images with an overlap rate of no less than 5%, which is superior to Harris, SURF and ORB algorithms in overall mosaic quality and time cost.