Aiming at the problem of low image stitching accuracy due to the difficulty of extracting strip features from multispectral images of filter array, a strip stitching algorithm for multispectral images of filter arrays is proposed. Firstly, the effective range of each band of the multispectral image is determined, and the homography matrix is calculated using the geographic coordinate information of the image vertices and the vertex coordinates to project the image. Secondly, Scale-invariant feature transform (SIFT) algorithm was used to extract matching points of projected images, and the mean value of the coordinate difference of matching points was calculated as the translation relation between images. Finally, the projection transformation is performed on the single band bands in turn, and the projected images are stitched with inter-image translation to obtain a large area single band image. Theoretical analysis and experimental results show that this method can effectively improve the the stitching accuracy of multispectral images of filter arrays and has a high image stitching speed.