Scribd
Upload
27 views9 pages

CS485/685 Midterm Exam Sample

This document is the syllabus for a computer vision course being taught in the spring of 2010. It outlines the topics that will be covered during the semester, including the midterm exam which covers key concepts in edge detection, interest point operators, Gaussian filtering, and coordinate transformations. The midterm contains both multiple choice and short answer questions testing students' understanding of these fundamental computer vision topics.

Uploaded by

Suresh Ramanujam
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
27 views9 pages

CS485/685 Midterm Exam Sample

This document is the syllabus for a computer vision course being taught in the spring of 2010. It outlines the topics that will be covered during the semester, including the midterm exam which covers key concepts in edge detection, interest point operators, Gaussian filtering, and coordinate transformations. The midterm contains both multiple choice and short answer questions testing students' understanding of these fundamental computer vision topics.

Uploaded by

Suresh Ramanujam
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

CS485/685 Computer Vision

Spring 2010 – Dr. George Bebis


Midterm Exam (1:00 PM - 2:15 PM)

Name:_______________________________________________
1. [40 points] To get credit, you must justify your answers!

T F There are four main steps in edge detection.

T F While we cannot estimate the direction of an edge using the Laplacian, we can still
estimate its strength.

T F Using lens allows us duplicate the pinhole geometry without having to use very small
apertures.

T F Convolving a Gaussian with itself yields another Gaussian with double the standard
deviation of the original Gaussian.

T F Radial distortion affects mostly the pixels close to the center of an image.
T F Focusing in the human eye is achieved by varying the distance between the lens and
the retina.

T F The rank of the matrix whose SVD is shown below, is 2.

T F Ramp edges can be better localized using the gradient than the Laplacian.

T F If an n x n matrix is diagonalizable, then its columns form a basis in Rn

T F Each point in Cartesian coordinates has a unique representation in Homogeneous


coordinates.
2. [30 points] Short answer questions.

(a) What is the difference between convolution and correlation? Under what condition is
convolution equivalent to correlation?

(b) Explain how the Moravec interest operator works. What are its main weaknesses?
(c) What is the separability property of Gaussian convolution? Why would one want a
convolution filter to be separable?

(d) Define the terms “depth of field” and “field of view”. What are the parameters that affect
them and how?
(e) What are the main steps of the Canny edge detector? Describe each step.

(f) In many applications, an image is smoothed by applying Gaussian filters of several sizes.
Why would one want to smooth an image using different parameters of the Gaussian?
.
3. [10 pts] Consider the vector (7, 3, 2); what are its coordinates in a new coordinate
system which has been obtained by applying the following transformations on the
original system: translation by (4, -3, 7) , followed by rotation around Z axis by 90o.
4. [10 points] Using SVD decomposition, we can compute more efficiently the solution of
Ax=b, where b≠0. Show how.
5. (a) [5 points] How do we show that a set of vectors x1,x2, …, xn forms a basis?

(b) [5 points] Assuming that x1,x2, …, xn form a basis, show how to find the expansion of
another vector y onto this basis. Is the expansion unique?
6. Graduate Students Only [20 points] How does the Harris interest operator improve the
Moravec interest operator? Derive the “auto-correlation” matrix (i.e., give the proof). What
are the properties of the auto-correlation matrix? How is it useful for detecting interest
points?

You might also like