Related Experiment Videos
Markovian and autoregressive clutter-noise models for a pattern-recognition Wiener filter.
Sovira Tan1, Rupert C D Young, Chris R Chatwin
1School of Engineering and Information Technology, Laser and Photonics Research Group, University of Sussex, Falmer, Brighton, UK.
Applied Optics
|November 21, 2002
Summary
New Markovian and autoregressive models significantly improve target detection filters by providing better clutter-noise estimates than traditional white-noise models in realistic scenarios.
Area of Science:
- Signal processing
- Pattern recognition
- Image analysis
Background:
- Modern target detection filters rely on accurate clutter-noise estimation for optimal performance.
- The white-noise model is a widely used but potentially suboptimal approach for clutter-noise estimation.
Purpose of the Study:
- To evaluate Markovian and autoregressive models as alternatives to the white-noise model for clutter-noise estimation in target detection.
- To assess the performance improvement offered by these new models in realistic clutter conditions.
Main Methods:
- Simulations using the Wiener filter were conducted.
- Real clutter scenes were utilized to test the proposed models.
- Performance was compared against the traditional white-noise model.
Main Results:
- Both Markovian and autoregressive models demonstrated significantly better performance than the white-noise model.
- The proposed models showed consistent and similar results across different types of real clutter scenes.
- Enhanced accuracy in clutter-noise estimation was observed.
Conclusions:
- Markovian and autoregressive models offer a substantial improvement over the white-noise model for target detection filters.
- These models provide a more robust and generalizable solution for clutter-noise estimation in diverse realistic environments.
- The findings suggest a shift towards more sophisticated statistical models for improved pattern recognition in complex scenes.