Accuracy and Resolution of Measurement Extraction for Out-of-Focus Targets in a Focal Plane Array

This article develops statistically optimal measurement extraction and resolution algorithms for out-of-focus (OOF) targets in focal plane array imagery. This is a real-world problem when two targets are close and nearly on the same line of sight from the camera, which is (automatically) focused at a distance in-between the two targets. Previous work introduced a maximum likelihood method combined with a full-frame matched filter and a linearized generalized likelihood ratio test (GLRT) for two-target resolution. These methods have been shown to provide accurate measurements and resolution decisions under ideal conditions involving bright, well-separated, in-focus targets. The present study extends the analysis to OOF target scenarios by modeling the effects of different distances from the sensor on target intensity and point spread function spread parameter. Performance is evaluated through Monte Carlo simulations on two scenarios of interest. Results demonstrate that OOF targets introduce model mismatches that degrade measurement accuracy, even at high signal-to-noise ratios. In addition, the OOF situation adversely affects the linearized GLRT's resolution performance, with the area under the receiver operating characteristic curve decreasing by up to 14% compared to the in-focus baseline for the scenarios considered.