Target Detection in Color Imagery: On the Path to a Color Target Acquisition Model

摘要

Existing target acquisition models are effective in the prediction of Target Acquisition (TA) performance for monochrome or single band imagers (visible or infrared). There is currently no performance model for color imagers or fused imagery rendered on color displays. This study is a first step in extending the intensity contrast-based TA models to a three-dimensional color space. The monochrome TA model was developed with a variable, perceived signal to noise ratio threshold (SNRT), in contrast space, to determine target detection probability. In this research, we determined the noise-limited SNRT in the chroma direction for three equally spaced hue angles of CIELAB (Commission Internationale Eclairage L* A* B*) color space. The comparison of the chroma SNRT in these three hues of L*, a*, b* space that may allow an extension of the noise-limited sensor model to this three-dimensional space for color display modeling. Such a model improvement would result in a sensor performance model for both color imagers and for fused imaging color displays. This study investigates the target detection signal to noise threshold of single targets in a noisy color background. Noise is described in terms of signals of both target and noise at three different hues. Target images are presented to observers who are required to locate the target and select it with a mouse. The format of each experiment is that of a "forced choice" human detection perceptual experiment. The images are 64 pixel square with targets of one or four pixels. The results are the signal to noise thresholds in the chroma direction for the three different hues.