Traditional correlation filtering methods produce classification results by processing one image frame at a time. The filters are designed to yield well defined correlation peaks when a pattern or object of interest is present in the input image. However, the decision process is memory-less, and does not take advantage of the history of results on previous frames in a sequence. Recently, Kerekes and Kumar introduced a new Bayesian approach for multi-frame correlation that first produces an estimate of the object's location based on previous results, and then builds up the hypothesis using both the current data as well as the historical estimate. In this paper, we examine a simple approximation to their approach which directly uses the correlation filter outputs while avoiding the need for density functions and explicit probability calculations. Preliminary analysis shows that the simplified approach has the potential for also yielding significant performance improvements over the conventional approach based on individual frames.
[1]
B. V. Vijaya Kumar,et al.
Unconstrained correlation filters.
,
1994,
Applied optics.
[2]
Abhijit Mahalanobis,et al.
Quadratic correlation filter design methodology for target detection and surveillance applications.
,
2004,
Applied optics.
[3]
Abhijit Mahalanobis,et al.
Distance classifier correlation filters
,
1994,
Defense, Security, and Sensing.
[4]
R. Kerekes,et al.
Enhanced Video-Based Target Detection using Multi-Frame Correlation Filtering
,
2009,
IEEE Transactions on Aerospace and Electronic Systems.
[5]
B V Kumar,et al.
Tutorial survey of composite filter designs for optical correlators.
,
1992,
Applied optics.