CS1674: Homework 5 - Written

1. [3 pts] List three reasons why line fitting (fitting a line to points) is not complete just by extracting edges (finding gradients over a threshold).
2. [3 pts] Imagine we are using k-means clustering to compress images by discarding some colors (replacing each color with the mean in its cluster). We also want to compute the error, i.e. the difference between the original image's intensities and the post-compression intensities. Which error would be larger: the one corresponding to k=5 or to k=15? Why?
3. [4 pts] List the steps required to create a mosaic using homography estimation.