Scale-Space and Feature Detectors
FAST, difference-of-Gaussian, and their relatives, judged by repeatability and invariance.
Key Terms
FASTDifference of GaussiansSIFT DetectorSURFRepeatabilityInvariance
Python Tools
OpenCVNumPyscikit-imageMatplotlib
Textbook Reference
Szeliski Ch.7
Topics Covered
- 1Gaussian scale-space and automatic scale selection
- 2FAST corner test and machine-learned corner acceleration
- 3Difference-of-Gaussians as a Laplacian approximation and SIFT keypoint detection
- 4SURF box filters and integral-image speedups
- 5Evaluating detectors by repeatability and invariance to scale, rotation, and illumination
Learning Outcomes
- Explain how scale-space enables scale-invariant keypoint detection
- Contrast FAST, DoG, and SURF detectors by mechanism and cost
- Measure detector quality using repeatability and invariance criteria
Lecture Infographics
1 image file
Lecture Infographics
Lecture PDFs
Scale-Space and Feature Detectors
The embedded PDF preview is disabled to reduce Vercel bandwidth usage. Open the lecture file directly in Google Drive for viewing, downloading, or presenting.