Abstract: The first two intrinsic mode functions of each image are obtained by using empirical mode decomposition algorithm to decompose images which need stitch. Through this method, the feature points are extracted and matched from the first two intrinsic mode function’s superimposition of each image. Based on these matched feature points pairs, parameters of the transformation matrix are calculated and then images are stitched. Experimental results have shown that this method effectively improves the accuracy rate of the feature points matched and has strong robustness. On image fusion, a novel weighed technique based on cosine relationship transform is proposed. With this technique, not only the seamless stitching images can be obtained, but also the phenomenon of ghosting in the overlapping stitching region can be eliminated.

Key words: computer application, image stitching, empirical mode decomposition, cosine relationship transform