I shot several pictures of my house from the same position but different angles.
In this section, I recovered the parameters of the transformation between each pair of images. I defined upto 14 correspondences, creating an over-determined sytem of linear equations. I used least squares to solve for homography matrix.
Using the inverse of the homography matrix, we can warp the image to a front-parallel image. I used cv2.remap to remap the images to new warped locations. Below, you can find examples of two images rectifies.
|
|
|
|
|
|
Finally, I tried to use linear blending and and laplacian stacks to stitch the images. I am still struggling with figurng out the bounding boxes for these images. Here are my current results.
 |
 |
 |
 |
The coolest thing I learnt was how we can take images from different angles and persepectives, project them onto one frame and amek panoramic images. I still need to figure out bouding boxes and make my blending better.