Related Experiment Videos
Point target detection and subpixel position estimation in optical imagery.
Vincent Samson1, Frédéric Champagnat, Jean-François Giovannelli
1Office National d'Etudes et de Recherches Aérospatiale, 29, Avenue de la Division Leclerc, 92322 Châtillon Cedex, France. vsamson@irisa.fr
Applied Optics
|January 23, 2004
Summary
This study introduces advanced image processing techniques for accurately detecting point objects and estimating their subpixel positions amidst background clutter. Improved methods enhance detection performance, overcoming limitations of traditional filters.
Area of Science:
- Image processing
- Computer vision
- Signal processing
Background:
- Distinguishing point objects from clutter is challenging.
- Object signatures can vary due to aliasing at subpixel locations.
- Conventional matched filters perform poorly in these conditions.
Purpose of the Study:
- To develop improved methods for point object detection and subpixel localization.
- To address the performance degradation caused by aliasing effects.
- To evaluate the impact of sensor design on detection and estimation.
Main Methods:
- Proposed alternative detectors: approximate and generalized likelihood-ratio tests.
- Compared performance against conventional pixel-matched filtering.
- Investigated two types of subpixel position estimators.
Main Results:
- Likelihood-ratio tests significantly outperform pixel-matched filtering.
- Demonstrated improvement in both detection and subpixel estimation.
- Highlighted the critical role of sensor design.
Conclusions:
- Novel detectors offer superior performance for point object detection in cluttered scenes.
- Subpixel estimation accuracy is influenced by aliasing and sensor characteristics.
- Sensor design is a key factor in optimizing imaging system performance.