Enhanced matched filter (EMF) comprises a distortion tolerant correlation filter and its associated threshold. It is an effective signal detection tool with superior immunity to noise, distortion, and clutter, used for imagery-based detection, authentication, classification, recognition, and tracking of targets of interest. The EMF is synthesized by combining multiple image templates of the target of interest, acquired under prescribed target states and view conditions. In autonomous vision and tracking systems, one EMF can potentially replace copious manifold of exemplar images without adversely affecting the classifier performance. This leads to proportional reduction of operation phase computational load, and concomitant smaller footprint, lighter, faster, and more power efficient smart vision systems. This paper develops the underlying theory of EMF and provides the analytical models of its operation. The EMF and the standard matched filter (MF) performance results based on analytical formulations and empirical studies are presented and are compared to the performance data using virtual and real test images. Results pertaining to the performance comparison of the EMF and the synthetic discriminant function (SDF) filters are also presented.
Enhanced Matched Filter, Correlation Filtering, Distortion Tolerant, Image Classification
Cite this paper
Kaveh Heidary. (2021) Enhanced Matched Filter Theory and Applications. International Journal of Signal Processing, 6, 39-52