Research On Digital Image Mosaic

Digital image mosaic is widely used in many fields. We compare algorithms of digital image stitching in this paper, and deeped in the researching of algorithm of image stitching base on scale-invariant keypoints features.Wether the descriptor of images for stitching is accurate determines there is a "ghosting" or not. We solved the "ghosting" fundamentally and got and more accurate match by adopting Lowe's methods.For exposure differences, we proposed an adjustment based on hist of images. As the two images for stitching, we estimate their hists and adjust these two images refer to the image that has a symmetrical gray scale. Pixels have different values according to their distance to the seam-line. After adjustment, there will be more natural transition between the seam-line.Base on algorithms above, we established a programmed system of digital image stitching base on invariant feature keypoints, including SIFT descriptors extracting module, SIFT descriptors reducing module and stitching module, and have got a better result.