DataTerminology.comNEW
Also known as: Speeded-Up Robust Features
A fast, robust local feature detector and descriptor using integral images and Hessian-based detection for scale- and rotation-invariant matching.
DESCRIPTION
SURF accelerates local feature detection and description, inspired by SIFT but optimized with integral images and approximations for speed. It detects keypoints with a Hessian-based measure and computes robust descriptors for matching across images. SURF is suited to real-time applications but was historically patent-encumbered, limiting open use. It remains a reference for efficient classical features and is implemented in libraries like OpenCV.
ALTERNATIVE DEFINITIONS
European Conference on Computer Vision
In this paper we present SURF, a novel scale- and rotation-invariant interest point detector and descriptor, which approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness.
Stresses invariance and performance compared to prior methods.
Wikipedia
Speeded Up Robust Features (SURF) is a robust local feature detector, first presented by Herbert Bay et al. in 2006.
Provides concise origin and purpose of SURF.
RELATED TERMS
Hessian Detector Integral Images Feature Matching RANSAC Descriptor OpenCV Scale Space
COMPETING TERMS
KNOWLEDGE AREAS
TARGET AUDIENCE
EXTERNAL SOURCES
- Wikipedia - Speeded Up Robust Features
- OpenCV - OpenCV tutorial on SURF
- Papers with Code - Method overview and benchmarks
TOP QUESTIONS
How does SURF achieve speed gains over SIFT?
It uses integral images and approximated filters for fast Hessian-based detection and efficient descriptor computation.
Category: Design
Is SURF free to use?
SURF was patent-encumbered, restricting use historically; check current licensing and alternatives like SIFT or ORB for open usage.
Category: Compliance
When should I prefer SURF over SIFT?
When speed is critical and licensing allows, SURF can be a good balance of accuracy and efficiency.
Category: Application
What are common SURF pitfalls?
Using default thresholds may yield too many or too few keypoints; tune parameters and apply geometric verification.
Category: Mistakes
How does SURF compare to ORB?
SURF is often more accurate but slower and historically encumbered; ORB is fast, binary, and license-friendly.
Category: Alternatives
What tasks suit SURF?
Real-time matching for tracking, stitching, and object recognition when classical methods are appropriate.
Category: Application
What is a hot topic around SURF today?
Evaluating classical vs learned local features under domain shifts and resource constraints.
Category: Topic
How to improve SURF matching robustness?
Use ratio tests, cross-check matching, and RANSAC-based model fitting.
Category: BestPractices
What is the biggest misunderstanding about SURF?
That it’s always faster than SIFT in practice; implementation and parameter choices strongly affect speed.
Category: Misunderstanding