Homework #1
EECS-841
Due: Friday, Sept 19 (midnight)
Total Points: 13
1. Thresholding and Edge Detection (5 points)
Although natural images are highly complex and cluttered, some applications of computer vision
allow the subject matter to be much more tightly controlled, resulting in simpler images. One
example is automated product inspection, where a factory's final products or components on an
assembly line may be moved across a light table to view them in silhouette. In these cases,
simply thresholding the image is often enough to locate object boundaries.
Better thresholding can be achieved by combining techniques from edge detection. One
approach is to use edge detection to choose the threshold values. In this problem, we will
explore that approach further.
a) (2 points) Write a Matlab function
function hist = hist2D(input_values)
that accepts a 2xN array (an array with two columns), and produces a 50x50 two-dimensional
histogram. Specifically, the program should
1) Compute the minimal and maximal values of each column of input_values
2) Normalize input_values so that the values of each column lie between 0 and 50. Let's call
this normalized array normed_input_values.
3) Construct the output histogram hist so that hist(i,j) (the ith column and jth row of
hist) contains an integer that is the number of rows r of normed_input_values such that
i-1 < normed_input_values(r,1) <= i
and
j-1 < normed_input_values(r,2) <= j
For example, if input_values was given by:
input_values = ![0, 0;
! 1, 1;
! 0.5, 0.5];
Then normed_input_values would be given by:
normed_input_values = ![ 0, 0;
! 50, 50;
! 25, 25];
and hist would be a 50x50 array, zero everywhere except
hist(50,50) = 1
hist(25,25) = 1
