The invention discloses a seamless splicing method for microscopic images with a low overlapping rate. The method includes the steps of selecting a reference image and regions of interest (ROIs) of images to be spliced, and pre-processing the ROIs; extracting SURF feature points of the ROIs, and roughly matching the feature points using an FLANN method; removing matching pairs with an over large spatial distance according to a deviation range of two images; conducting statistics of discrete distribution of spatial distances of the remaining matching pairs, taking an interval corresponding to the maximum distribution probability as a standard range, further removing erroneous matching pairs, and if the matching pairs are insufficient, returning back to the previous step to modify feature extracting parameters and matching parameters to obtain more matching pairs; calculating a positional relation between two images according to the remaining matching pairs; and adjusting the brightness of the images to be spliced, and fusing the images using a weighted average method. The method is based on an improved SURF feature point matching method, can effectively improve the feature matching accuracy and efficiency under a low overlapping rate of the images, and thus improves the quality and speed of the image splicing.