Assignment #1
Due Thursday, Sept. 19 by midnight
Implement the DLT homography estimation discussed in class on Sept. 10
(and detailed in the handout) in Matlab.
Rectify these three images based on the correspondences between the
corners of a square under perspective in each image and a hypothetical,
frontally-viewed square.
Experiment with using more than four correspondences, differently-sized squares
in the image, differently-sized target squares, and rectangular image regions
(mapped to rectangular targets, of course). How do the rectified images vary
for each scene?
Turn in three rectified images (your best representative of each sample, with
data on correspondence points used for each and the dimensions of the target
square/rectangle), all Matlab code, and a short description of your observations,
including how you handled the target image size and unmapped areas in it.
Notes
Use ginput for mouse clicks in Matlab to record correspondences.
Here's a stub of code for reading in an image, displaying it, and picking
points: pick_points.m