Related Experiment Video
Updated: Jun 19, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
Adaptive feature-specific imaging for recognition of non-Gaussian classes.
Pawan K Baheti1, Jun Ke, Mark A Neifeld
1Department of Electrical and Computer Engineering, 1230 East Speedway Boulevard, University of Arizona, Tucson, Arizona 85721, USA. baheti@email.arizona.edu
Applied Optics
|October 3, 2009
Summary
This study introduces an adaptive feature-specific imaging (AFSI) system for M-class recognition. The AFSI system significantly reduces measurement needs, especially in low signal-to-noise ratio (SNR) conditions, outperforming conventional adaptive systems.
Area of Science:
- Signal Processing
- Machine Learning
- Pattern Recognition
Background:
- Adaptive systems are crucial for efficient data processing in recognition tasks.
- Accurate density estimation is vital for classification performance, especially with non-Gaussian distributions.
Purpose of the Study:
- To introduce and evaluate an Adaptive Feature-Specific Imaging (AFSI) system for M-class recognition.
- To compare the measurement efficiency of AFSI against an Adaptive-Conventional (ACONV) system.
Main Methods:
- Utilized nearest-neighbor-based density estimation for class-conditional densities.
- Employed adaptive projection basis and sequential hypothesis testing for decision-making.
- Refined density estimates and projection basis using prior measurement data.
Main Results:
- AFSI demonstrated significant performance improvements over ACONV, particularly at low signal-to-noise ratios (SNR).
- For M=4 hypotheses and SNR=-10 dB, AFSI required 30 times fewer measurements than ACONV for a P(e)=10(-2).
- Experimental validation confirmed the efficacy of the proposed AFSI system.
Conclusions:
- The AFSI system offers a substantial reduction in required measurements for M-class recognition tasks.
- Adaptive feature selection and sequential testing are key to AFSI's efficiency gains.
- AFSI is a promising approach for scenarios demanding high efficiency in low SNR environments.
