Advanced image processing techniques for extracting regions of interest using multimode IR processing

Abstract

As large format single color and multicolor sensors proliferate challenges in viewing and taking action on a much higher data volume becomes challenging to end users. We report on processing techniques to effectively extract targets utilizing multiple processing modes — both multiple spectral band and multiple pre-processed data bands from Focal Plane Array (FPA) sensors. We have developed image processing techniques which address many of the key pre-processing requirements, including scene based non-uniformity correction of static and dynamic pixels, multiband processing for object detection, and reduction and management of clutter and non-targets in a cluttered environment. Key motivations for these techniques include image pre-processing extracting small percentages of the image set with potentially high likelihood targets and then transmitting “active” pixel data while ignoring unchanging pixels. These techniques have demonstrated significant reductions in the raw data, and allow the end user to more intelligently select potential data types for object identification without requiring a person in the loop.