Metric for Analyzing Radar Image Effectiveness (MARIE)

The challenge is to develop an image quality metric and prediction model for SAR imagery that addresses exploitation by automated tools (i.e., ATR) and trained imagery analysts. Our approach rests on a deep understanding and experience with the Radar National Imagery Interpretability Ratings Scale (Radar NIIRS) and the previous development of the General Image Quality Equations (GIQE) for other modalities, while incorporating experience in ATR characterization and performance modeling. Phase 1 establishes the theoretical baseline by addressing three fundamental issues. We will clarify the relationship between human perception of image quality (i.e., NIIRS) and machine-centric measures of image quality (i.e., ATR performance). The modeling and analysis will identify the key factors (sensor and acquisition parameters) that affect image quality for ATR. Lastly, we will assess the effects of diverse background clutter environments and their role in the formulation of a prediction model. By developing a clear understanding of these three issues in Phase 1, the Phase 2 effort can focus on the rigorous, quantitative analysis necessary to produce a robust image quality metric and prediction model. BENEFIT: We expect the MARIE technology to have immediate and tangible benefit for a number of military sensor and image exploitation systems. In particular, MARIE will help the ATR community develop ATRs with well-defined performance characteristics as a function of SAR sensor characteristics. Likewise, the SAR sensor development community will benefit from the ability to conduct design and trade studies using both an image quality prediction model and an image quality metric which have proven correlation with ATR performance. This will clearly be relevant to the various Performance Driven Sensing initiatives in the Air Force, as well as missile seeker design programs and persistent ISR programs across all branches of the military.