As the unmanned aerial vehicle (UAV) image has large data amounts and high resolution, the stitching error cannot be ignored during the image mosaic process which involves multiple images. Furthermore, with the number of mosaic images increasing the errors will continue to accumulate. The main error sources of UAV image mosaics are analyzed. A deep study is made on the image overlap and the error accumulation problem. The optimal overlap of UAV image mosaic is determined, and the original image is sparsed according to the optimal overlap. Aiming at the error accumulation problem during splicing process, the block splicing method is proposed. The experimental results show that the number of images can be reduced according to the optimal overlap, and the stitching error can be reduced effectively as the block splicing method is adopted to splice sparsed image. The vision effect of image splicing can be improved, with the efficiency of image processing enhanced in a certain extent.