Digital Visual Effect

Image Stitching - Panorama

We take images by rotating a camera on a tripod. To stitch the images together, we warp the images to the cylindrical coordinate first. We then extract the feature points by Harris corner detection method [1] and apply the SIFT feature description [2] to each points. The Ransac algorithm is applied to compute the pairwise alighment according to the points associated with SIFT feature description. At last, we stitch and blend the images into a panorama given the alignment parameters.

Results:

Data Set 1:
input set (.zip), constructed panorama:

(This artifact won the second place among the class.)

Data Set 2 (Testing Sample):
input set (.zip), constructed panorama:

References:

  1. Chris Harris, Mike Stephens, "A Combined Corner and Edge Detector," 4th Alvey Vision Conference, 1988, pp147-151.
  2. David G. Lowe, "Distinctive Image Features from Scale-Invariant Keypoints," International Journal of Computer Vision, 60(2), 2004, pp91-110.
  3. Richard Szeliski, "Image Alignment and Stitching," unpublished draft, 2005.
  4. R. Szeliski and H.-Y. Shum, "Creating full view panoramic image mosaics and texture-mapped models," SIGGRAPH 1997, pp251-258.
  5. M. Brown, D. G. Lowe, "Recognising Panoramas," ICCV 2003.