Related Experiment Video
Updated: Jul 7, 2026
03:31
End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
Published on: December 15, 2023
Performance assessment of feature detection algorithms: a methodology and case study on corner detectors
1Department of Electronic and Electrical Engineering, University of Sheffield, Sheffield S1 3JD, UK. p.rockett@ shef.ac.uk
Summary
This study introduces a new method to evaluate feature detector labeling performance. It assesses three corner detectors, highlighting their limitations in distinguishing corners from non-corner features.
Area of Science:
- Computer Vision
- Image Processing
- Pattern Recognition
Background:
- Feature detection is crucial for image analysis.
- Evaluating the accuracy of feature detectors, especially corner detectors, is essential.
- Existing methods may not comprehensively assess labeling performance.
Purpose of the Study:
- To propose a generic methodology for evaluating the labeling performance of feature detectors.
- To assess the performance of three prominent corner detectors using this methodology.
- To identify labeling deficiencies related to feature discrimination.
Main Methods:
- Development of a generic methodology for performance evaluation.
- Creation of a dedicated test set for evaluation.
- Application of the methodology to Kitchen-Rosenfeld, Paler et al., and Harris-Stephens corner detectors.
Main Results:
- The methodology provides a framework for quantitative assessment.
- Labeling deficiencies were identified in the evaluated corner detectors.
- Performance issues were linked to the detectors' ability to discriminate corners from non-corners.
Conclusions:
- The proposed methodology offers a robust approach to evaluating feature detector labeling.
- Understanding detector limitations in feature discrimination is key for improved performance.
- This work contributes to the advancement of reliable feature detection techniques.