PART 1 COMING IN HOT
1.1: Shooting Pictures
The ones I'll use for panoramas are currently on my phone (to be revealed). For now let's focus on image rectification!
1.2: Recovering Homographies
I first had to find the 3x3 homography that relates one image to the next, in order to warp it. To do this, I used least squares to recover 8 values from 18 points that I recorded between both images (8 per image). The 9th part of the homography is a scaling factor that I set to 1.
1.3: Rectification (where we stop for today)
Let's set some images right. I take the inverse of the homography matrix and apply it to the coordinates of my image in order to receive new coordinates corresponding to the warped image. I then sample the old image for color values in order to reconstruct it. Here are my results! (I found everything in my kitchen because I haven't left this place in a week). The results are a bit rough due mainly to me eyeballing the point selection (I eyeballed a square on the cup cover, mapping it to another eyeballed square on a black screen...so there is some noise).
|
|
|
|