CS 766 - Computer Vision
- Midterm Exam: Thursday November 9 from 11:00 a.m. - 12:30 p.m. in class (note late ending time)
- Covers readings up to that point in the semester, including
Chapters 1.1, 2, 3.1.1, 3.2, 7 (except 7.3),
8, 10.1, 11, 14.1, 14.4, 14.5, 18.2.2, 22.1, and 22.2.
Includes the notes on projection by Cipolla and Gee,
notes on edge detection by Huttenlocher, notes by Cipolla and Gee on stereo vision,
SIFT tutorial, the paper by Szeliski, "Video mosaics for virtual environments,"
the paper by Comaniciu and Meer on the mean shift algorithm,
the notes by Sonka et al. on snakes, and the notes by Sethian on level sets and the fast marching
method. You are not responsible for the
paper by Dyer on volumetric scene reconstruction, the paper by Szeliski and Shum,
the paper by Heckbert, the paper by Dimitrov et al., and the following parts of the assigned
reading in the textbook: 1.2 - 1.5, 4, 8.1, 14.2, and 14.3.
Covers material and readings associated with Homeworks 1, 2 and 3.
Covers lectures and lecture
notes for material covered in class through November 7.
- Most of the main topics covered: Marr paradigm, camera models (projective, perspective,
orthographic, weak perspective, affine), homogeneous coordinates, rigid body
transformation, camera calibration matrix (K), vanishing points, vanishing
(aka horizon) lines, properties of camera projection models, intrinsic and
extrinsic camera parameters, camera calibration problem,
Shannon sampling theorem, Nyquist rate,
aliasing, false contours, digitization: sampling and quantization,
Cipolla and Gee projection handout,
segmentation problem, segmentation by
thresholding (mode method, p-tile method, adaptive thresholding, recursive
thresholding), digital geometry (neighbors, adjacency, connected components,
border, hole, etc.), connected component labeling,
medial axis transform (MAT), distance transform (DT), thinning,
quadtrees, digital metrics,
k-means segmentation, mean shift segmentation,
normalized cut segmentation, snakes (active contours),
level sets and fast marching for segmentation,
causes of intensity edges, edge detection goals, edge operator
properties (linear, separable, etc.), 1st derivative edge operators (gradient,
Roberts, Prewitt, Sobel), smoothing (mean, median, Gaussian), convolution,
non-maxima suppression, Canny edge operator, hysteresis thresholding,
2nd derivative edge operators (Laplacian, Marr-Hildreth, DoG),
scale space, pyramids (Gaussian, Laplacian, oriented),
corner detection (Harris, and Tomasi and Kanade), SIFT keypoint detector, SIFT descriptor,
Hough transform, mosaics, cross-correlation matching,
sum-of-squared-difference (SSD) matching,
snakes, level sets,
epipolar geometry, fundamental matrix, essential matrix, constraints for stereo
correspondence, rectification, and multi-baseline stereo.
- Exam 1 Solution
- Format: The exam will cover topics up through that time,
including readings in the textbook, papers, and
homework assignments. You may bring into each exam one (1) 8.5" x 11"
sheet of paper with any notes you want on both sides. Bring a calculator
to the exam. The exam will
focus on main ideas and algorithms, not derivations or proofs. See old exams below for
the types of questions that will be asked.
- Old Exams