Article ID Journal Published Year Pages File Type
4969169 Information Fusion 2018 18 Pages PDF
Abstract

•An eye movement-based methodology for quantitatively evaluating fused images.•Measures of central and peripheral visual processing efficiency.•Performance measures differentiated fused images from component single band images.•Performance measures differentiated two-way fusion schemes and fusion methods.•Conducted stimulus modeling in order to predict performance outcomes.

Human performance measures were used to evaluate the perceptual processing efficiency of infrared and fused-infrared images.  In two experiments, eye movements were recorded while subjects searched for and identified human targets in forested scenes presented on a computer monitor.  The scenes were photographed simultaneously using short-wave infrared (SWIR), long-wave infrared (LWIR), and visible (VIS) spectrum cameras. Fused images were created through two-way combinations of these single-band images.  In Experiment 1 the single band sensors were contrasted with a simple average fusion scheme (SWIR/LWIR). Analysis of subjects' eye movements revealed differences between sensors in measures of central processing (gaze duration, response accuracy) and peripheral selection (detection interval, saccade amplitude). In Experiment 2 this methodology was applied to compare three two-way combinations of sensors (SWIR/LWIR, SWIR/VIS, VIS/LWIR), produced by state-of-the-art fusion methods.  Peripheral selection for fused images tended to exhibit a compromise between the performance levels of component sensor images, while measures of central processing showed evidence that fused images matched or exceeded the performance level of component single-band sensor images. Stimulus analysis was conducted to link measures of central and peripheral processing efficiency to image characteristics (e.g. target contrast, target-background contrast), and these image characteristics were able to account for a moderate amount of the variance in the performance across fusion conditions. These findings demonstrate the utility of eye movement measures for evaluating the perceptual efficiency of fused imagery.

Related Topics
Physical Sciences and Engineering Computer Science Computer Vision and Pattern Recognition
Authors
, , , ,