CSO Resolution

Integrated Sensors, Incorporated (ISI) proposes to develop a frequency-domain closely spaced object (CSO) condition detection algorithm and evaluate a variety of deconvolution approaches. We propose a combined CSO / deconvolution process where information extracted from the CSO detection process will drive the design of the deconvolution filters. We will also use the CSO information to characterize the spacing geometry and sizes of the CSO's. ISI has already developed numerous deconvolution algorithms under previous efforts. Algorithms such as adaptive Wiener filters and recursive maximum likelihood / entropy optimizations are among those already available in MATLAB. All of these require precise models of the system MTF (Modulation Transfer Function), which ISI has developed under two previous efforts. The Wiener approach is nonrecursive and would be the simplest to implement, while the recursive approaches tend to perform better under nongaussian conditions. A major challenge is address the overshooting issue of the recursive approaches. The Cramer-Rao Lower Bound (CRLB) provides an ideal framework from which to address this issue. Innovation is also required to structure these algorithms for an efficient, real-time solution. ISI will include this technology as part of its ImageExpressT commercial motion analysis system, to enhance measurement accuracy.