Aiming at the problem of low registration accuracy in the mosaic of oblique remote sensing images and ghosting in overlapping areas, a mosaic method using triangulation for small bin registration and weighted fusion is proposed. First, solve the global homography matrix for image pre-alignment and use the transformation matrix to extract the image overlap area, then use feature matching to construct the overlap area Delaunay triangulation and perform affine transformation on the corresponding triangulation one by one to achieve precise registration. Finally, the fade - in and fade - out fusion is used to eliminate the shadow of the overlapping area and improve the visual effect of image mosaic. The quality of image registration and mosaic was evaluated by combining subjective evaluation (CCIR500-1 subjective evaluation standard) with objective evaluation (PSNR evaluation and SSIM evaluation). Experimental results show that this method can effectively improve the registration accuracy and eliminate the shadow of the overlapping area of image mosaic. Compared with the current main algorithms, this algorithm has higher registration accuracy and better visual effect of image mosaic. Therefore, it can be effectively applied to the field of oblique image mosaic.