Methodology For Evaluating Statistically Predicted Versus Measured Imagery 1 Rob Kooper, 1 Peter Bajcsy and 2 Dennis Andersh 1 National Center for Supercomputing Applications (NCSA) University of Illinois at Urbana-Champaign, IL and 2 Scientific Applications International Corporation (SAIC), Champaign, IL Email: kooper@ncsa.uiuc.edu , pbajcsy@ncsa.uiuc.edu , dennis.j.andersh@SAIC.COM ABSTRACT We present a novel methodology for evaluating statistically predicted versus measured multi-modal imagery, such as Synthetic Aperture Radar (SAR), Electro-Optical (EO), Multi-Spectral (MS) and Hyper-Spectral (HS) modalities. While several scene modeling approaches have been proposed in the past for multi-modal image predictions, the problem of evaluating synthetic and measured images has remained an open issue. Although analytical prediction models would be appropriate for accuracy evaluations of man-made objects, for example, SAR target modeling based on Xpatch, the analytical models cannot be applied to prediction evaluation of natural scenes because of their randomness and high geometrical complexity imaged by any of the aforementioned sensor modality. Thus, statistical prediction models are frequently chosen as more appropriate scene modeling approaches and there is a need to evaluate the accuracy of statistically predicted versus measured imagery. This problem poses challenges in terms of selecting quantitative and qualitative evaluation techniques, and establishing a methodology for systematic comparisons of synthetic and measured images. In this work, we demonstrate clutter accuracy evaluations for modified measured and predicted synthetic images with statistically modeled clutter. We show experimental results for color (red, green and blue) and HS imaging modalities, and for statistical clutter models using Johnson’s family of probability distribution functions (PDFs). The methodology includes several evaluation techniques for comparing image samples and their similarity, image histograms, statistical central moments, and estimated probability distribution functions (PDFs). Particularly, we assess correlation, histogram, chi-squared, pixel and PDF parameter based error metrics quantitatively, and relate them to a human visual perception of predicted image quality. The work is directly applicable to multi-sensor phenomenology modeling for exploitation, recognition and identification. Keywords: synthetic image evaluation, statistical multi-sensor phenomenology modeling, hyperspectral imagery. 1 INTRODUCTION The problem of multi-sensor phenomenology modeling for exploitation, recognition and identification includes several fundamental theoretical, experimental and validation issues. In this paper, we focus only on validation issues in the context of multi-sensor phenomenology modeling. The objective of multi-sensor modeling, as defined in this work, is to predict image appearance (or pixel values) based on (a) previous knowledge about viewed scene and objects, (b) existing data (saved measurements), and (c) developed prediction models. The goal of a validation component in the process of multi-sensor modeling is to assess the goodness of image predictions based on modeling or application criteria. In order to design a validation methodology, one has to understand other components of the multi-sensor phenomenology modeling, such as data dimensionality and type, sensor modality and prediction models. In our work, we consider imaging sensors that form a 2D raster image (or a grid of measurements). The imaging modalities could include several known sensors, such as Synthetic Aperture Radar (SAR), Electro-Optical (EO), Multi-Spectral (MS) and Hyper- Spectral (HS) modalities. We primarily focus on color (red, green and blue denoted as RGB) and HS imaging modalities. While several scene modeling approaches have been proposed in the past for single-modal image predictions 1, 3, 6, 7, 8 , the problem of evaluating synthetic and measured images has remained an open issue. Although analytical prediction models would be appropriate for accuracy evaluations of object models, for example, SAR target modeling based on Xpatch 2 or thermal IR target modeling based on MuSES 4 , the analytical models cannot be applied to prediction