CSE152 – Intro to Computer Vision – Assignment #2
Instructor: Prof. David Kriegman.
http://cseweb.ucsd.edu/classes/sp12/cse152-a
Due Date: Thu. May 11, 2012.
Instructions:
•Attempt all questions
•Submit your assignment electronically by email to [email protected] with the subject line
CSE152 Assignment 2. The email should include 1) A 1) a written report that includes all
necessary output images and written answers (e.g., as a Word file or PDF), and 2) all necessary
Matlab code. Attach your code and report as a zip or tar file.
1 Binarization:
Write a Matlab function to implement the ’peakiness’ detection algorithm described in class. This
algorithm should automatically determine an intensity level to threshold an image to segment out
the foreground from the background. The output of your function should be a binary image that is 0
for all background pixels and 1 for all foreground pixels. Apply this function to the image coins.png
and turn in the output image in your report. [9 points]
Notes:
•Load in an image as a grayscale, double-precision image, e.g.
img = im2double(rgb2gray(imread(’coins.png’)));
•You can use the Matlab function hist(img(:),numBins) to create a histogram of pixel intensities
as a first step to the peakiness detection algorithm. Using numBins=10 and a minimum distance
of 3 bins between peaks should work.
2 Morphological Image Processing
Morphology refers to application of set operations such as union and intersection on an image.
Morphological algorithms take as input an image and a structuring element, which encodes the
shape characteristics based on which the input image must be processed. Usually, the structuring
element is a (2k+ 1) ×(2k+ 1) array, where kis a small number (say, between 1 and 5).
In general, a morphological operation is performed by making a copy of the input image. The
structuring element, S, is compared to a (2k+ 1) ×(2k+ 1) window neighborhood, W, of each pixel
of the input image. If Wexactly matches S, then the corresponding pixel of the copy image is set
to a particular value (0 or 1, depending on particular operation), else that pixel in the copy is left
unchanged.
(a) Erosion: The desired effect in erosion is that any foreground pixel which has a background
pixel as a neighbor is set to background. Consider the 3 ×3 structuring element:
S1=
111
111
111
The 3 ×3 neighborhood of each pixel in the input image is compared to this structuring element
S1and the corresponding pixel in the copy image is set to background if they do not match, else
it is left unchanged. Thus, it causes the boundary of the foreground to shrink. Write Matlab
1
