Automated object detection with a correlation filter designed from a noisy image

Correlation filters used for object detection are commonly designed using explicit knowledge of the target appearance and the target shape. This assumption requires that the image of a target used for the filter design be manually processed. In this paper, we suppose that the target is given at unknown coordinates in a reference image corrupted by additive noise. Optimum correlation filter with respect to the peak-to-output energy ratio for object detection is derived. Computer simulation results are presented comparing the performance of the proposed filter with that of common correlation filters.