Guide-linked quiz · Medium Level
Homography and Image Stitching Quiz
Learn where homographies apply, how stitching pipelines align and blend images, and why parallax defeats a single global transform.
Question 1 of 8
How does a homography map points between image views?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- How does a homography map points between image views?
- A 3D rigid-motion transform between camera centers
- A projective mapping between image views of a plane
- A depth-dependent warp valid for every point in a translated scene
- A color correction between exposure values
- For a standard projective transform, how many non-collinear point pairs are the minimum for estimation?
- Three non-collinear point pairs
- Four collinear point pairs
- Eight arbitrary point pairs are always required
- Four
- When can a homography relate views of a general scene under camera motion?
- When the camera undergoes pure rotation about its center
- When a scene is planar, translation always makes the homography invalid
- Whenever a camera translates through a scene at mixed depths
- Only when the scene has no overlapping regions
- Why can camera translation create parallax that one global homography cannot remove?
- All depth layers move by the same image displacement
- The transform changes color but leaves position fixed
- Points at different depths shift by different amounts
- Parallax occurs only when the images have no overlap
- Which sequence is part of a full panorama stitching pipeline?
- Detect features, classify objects, and copy the first image over the second
- Feature matching, camera estimation, warping, seam selection, and blending
- Corner detection followed by semantic labeling only
- Exposure compensation without estimating image alignment
- Why can a photographed paper page often be rectified with a homography?
- Its printed text supplies metric depth at each corner
- The phone depth sensor is required to locate page corners
- A page can be rectified only when the camera rotates without translation
- The page is approximately planar
- What can happen when a moving person appears in overlapping panorama frames?
- The person may appear doubled or leave a visible seam
- The person becomes blurred because all pixels lose color
- The homography recovers the person's path between frames
- Moving people become the camera's rotation center
- What should be inspected after estimating a homography from feature matches?
- Only whether the transform matrix has nonzero entries
- Whether every match is an inlier, regardless of where points lie
- Whether inliers cover the overlap, reprojection residuals are small, and the warped images align plausibly
- Whether both images contain identical camera metadata
Keep testing yourself
More quizzes picked for your level.