Uses image mosaicing to generate a panorama from five pictures of the same scene.
 |
|---|
| Original image 1 |
 |
|---|
| Original image 2 |
 |
|---|
| Original image 3 |
First, SIFT is used to identify interest points.
 |
|---|
| Interest points for image 1 using SIFT |
NCC is used to establish correspondences.
 |
|---|
| Correspondences between images 1 and 2 using NCC |
Then, automated homography estimation is implemented using RANSAC with a linear least squares algorithm.
 |
|---|
| Inlier correspondences between images 1 and 2 using RANSAC |
 |
|---|
| Outlier correspondences between images 1 and 2 using RANSAC |
 |
|---|
| Homography mapping image 1 to 2 using RANSAC |
The homography is refined using a nonlinear least squares algorithm. Matrix multiplication is used to put all images in the same reference frame (that of the center image).
 |
|---|
| Homography mapping image 1 to 3 using the Levenberg–Marquardt algorithm |
Finally, all images are plotted on the same canvas to create a panorama.
 |
|---|
| Final panorama |