- Matching in 2D
- Be able to explain the difference bewteen symbolic and geometric matching or to decide for given methods, whether they are symbolic, geometric, or hybrid (ie. both).
- Be able to use the formalism for consistent labeling to describe a symbolic matching problem.
- Be able to determine if a given mapping is a consistent labeling and if not, what relationships are not satisfied.
- Be able to draw an interpretation tree that shows the search for a consistent labeling in a given problem, with dead ends (X) when some consistency check fails.
- Be able to explain how discrete relaxation differs from tree search.
- Be able to compute relational distance of a given mapping.
- Be able to explain how a tree search for computing relational distance would differ from the tree search for a fully consistent labeling.
- Be able to show how the local-feature focus method would work on a given example.
- Be able to explain the differences between local-feature focus, pose clustering, and geometric hashing.
- Be able to discuss what is the SAME in local-feature focus, pose clustering, and geometric hashing.

- Motion and Optical Flow
- Be able to answer questions about the following basic concepts in motion
- change detection by image subtraction
- meaning of optical flow
- detection of interest points
- matching of interest points
- MPEG compression (independent frames, predicted frames, between frames)
- standard optical flow assumptions and what might violate them
- histogram-based video segmentation

- Be able to answer (simple) questions about how the Lukas-Kanade procedure computes optical flow in a K x K window.
- Be familiar with the circumstances under which Lukas-Kanade fails to find the optical flow.

- Be able to answer questions about the following basic concepts in motion
- 3D Imaging, 3D Sensing, 3D Reconstruction
- Be able to compute values for image coordinates given camera coordinates and focal length.
- Be able to solve for the 3D point (x,y,z) given a pair of corresponding points (xl,yl) and (xr,yr), the baseline distance b, and the focal length f in the simple stereo model with parallal optic axes.
- Be able to apply correlation or symbolic matching to find some correspondences.
- Be able to explain the epipolar and ordering constraints and to draw pictures illustrating them.
- Be able to explain the idea behind structured light and how it differs from passive 2-camera stereo.
- Be able to answer questions about the 5 different coordinate systems needed to develop the camera model.
- Be able to discuss the difference between extrinsic and intrinsic camera parameters and answer questiona about these parameters.
- Be able to answer questions about the Tsai calibration procedure up to the level of detail in the book.
- Be able to explain why we don't just solve the 4 linear equations in 3 unknowns that come about due to a point correspondence in general 2-camera stereo.
- Be able to answer questions about how voxel carving works for 3D reconstruction.
- Be able to answer questions about the contour tracking process in the blood vessel reconstruction procedure.

- 3D Models and Matching
- Be able to decide which types of 3D models are suitable for given applications.
- Be able to discuss how 3D-3D and 3D-2D alignments are similar to and how they are different from 2D-2D alignment (ch 11).
- Be able to answer questions about the construction of spin images at each point of a mesh model of a 3D object.
- Be able to answer questions about how spin images are used in matching.
- Be able to answer questions about relational matching in the RIO system (ie. 2-graphs, hashing, retrieving models, voting, etc).
- Be able to explain the idea of functional object recognition and to make a case for using it or not using it for some given application.
- Be able to answer questions about what principal components does (not how, just what), eigenfaces, and appearance-based object recognition using these concepts.