Description
1) Compute and display the Harris pixel-wise cornerness function R values for the
image checker.jpg using a) Gaussian window/weighting function with a standard
deviation of σI = 1 (use 3σ mask size), b) Gaussian Gx,Gy gradients with a standard
deviation of σD = 0.7 (use 3σ mask size), and c) trace weighting factor of α = 0.05.
Give the values of R(17:23, 17:23) in your report. Next remove the smaller and
negative values in R (anything < 1,000,000). Display the thresholded R using
imagesc. Lastly, do non-maximum suppression on R (for this version, keep a location
only if a unique maximum is found in its 3×3 region) to identify the actual corner
points and display them on the original image. (Note: use double() and not
im2double() [as it scales values to 0-1] on checker.jpg) [5 pts]
2) Implement the FAST feature point detector using a radius of r = 3 (you can hardcode
the particular circle border locations), intensity threshold of T = 10, and a consecutive
number of points threshold of n* = 9. Run the detector on the image tower.png.
Display the image and overlay the FAST feature points. Repeat with T = {20, 30, 50}
and compare all four results. [6 pts]
figure;
imshow(tower);
hold on;
plot(fastX,fastY,’r.’);
hold off;
3) As usual, turn in and upload your material.
