Non-collinear TDI-CCDs sensor was introduced and its imaging characteristics were analyzed. An information
mending algorithm was adopted to stitch TDI-CCDs imagery. Key points of the algorithm are image matching and
modeling of coordinate transformation between two images. Particularly, piecewise polynomial transformation model
was adopted to describe the shift, rotation and the zoom factor of TDI-CCDs imagery after extracting tie points using
SIFT algorithm. Coefficients of polynomial transformation were obtained through the solution of error equation
constructed by large numbers of tie points and finally TDI-CCDs imagery was geometrically stitched. A certain high
resolution satellite remotely sensed imagery were adopted to verify this algorithm. The experiments show that total
precision of this information mending stitching algorithm reaches sub-pixel level.