The invention relates to an automatic registering and fused splicing method of multiple images. The unordered images are enhanced in a multi-dimension Retinex method; the images match each other by utilizing a logarithm polar coordinate transformation improved phase correlation algorithm, zooming, rotation and translation parameters are calculated, an overlapped area among the images is estimated roughly, and an ordering rule is formulated according to energy peak values of an impulse function; an SUFR algorithm is used to extract characteristic points from the overlapped area to be registered, a random sampling consistent algorithm is improved to purify matching point pairs accurately, model parameters are optimized, a transformation matrix among the images is established, and sequential images are spliced successively; and an NSCT algorithm is utilized, a fusion strategy is made to further fuse spliced images, and a spliced image after fusion is output. The method of the invention can be used to reduce influence of difficult manual ordering, fuzzy image details, low resolution, high noise, nonuniform illumination and the like on image registration and splicing, provides help for reducing cost of imaging equipment and accurate diagnosis of medical staff, and has good application prospects in the medical image assisted field.