Development of Phase-Only Filters for Sensor Imagery
Abstract
We used the statistical technique of factor analysis to design binary and ternary optical correlation filters to identify objects in the presence of unknown or non-repeatable distortions. We considered values of spatial frequencies of training set imagery as features and used those spatial frequencies in our filter depending on their variation across a training set. In addition, we provided general expressions for performance measures as a function of training set imagery. The potential of our approach was evaluated with infrared sensor imagery that varied in an unknown way. Our statistically designed filters were easily calculated and performed well in the presence of noise. Furthermore, the performance of our filters were varied by allowing for trade-offs in performance measures. Our filters reduced the sensitivity of binary and ternary phase-only filters to changes in an object's appearance when the input imagery varied in an unknown manner. Optical filter design, Optical correlation, Optical pattern recognition.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1993
- Accession Number
- ADA276722
Entities
People
- Samuel P. Kozaitis
Organizations
- Florida Institute of Technology