IMAGE WARPING and MOSAICING
1. Shoot the Pictures
I took these three images in my apartment's courtyard. In order to make mosaicing easier, I overlapped 50% of the images. I also made sure to rotate the angle without changing the position of shooting point.
|
|
|
2. Recover Homographies
I first calculated a projection matrix H by solving a system of linear equations. Since the degree of freedom for H is 8, we only need 4 data points to calcualte it. To make the result more accurate, I chose 7 points depicted below to calcualte the projective transformation. Since there are noise in the data, I used least square equation to calculate H instead of directly solving it.
|
|
3. Warp the Images
After calculating a projective matrix H, I transformed image1 into image2 using cv2.remap(). Although a part of it is cutted off, the resulting image is successfully projected into image2's perspective.
|