Target visibility index

Abstract

Quantitative description of the quality of images from the military sight-seeing systems, particularly those of armored vehicles, implies specific approaches different from those used in traditional image processing for civilian needs. Our analysis of the literature shows that there are several disadvantages in the definitions of the quantitative image quality indices, which makes them inapplicable for the characterization of targets in images from military sight-seeing systems. First, quantitative indices describing the quality of images introduced for civilian applications and techniques for their measurements are not target-oriented. In most cases, image quality indices available in the literature characterize the image as a whole but turn out to be irrelevant to the visibility and conspicuity of a target. High image quality indices do not guarantee high visibility and conspicuity of a target. Contrarily, frequently high contrast and thereby visibility of a target is accompanied by abnormal (enormously high or low) lighting of the background, target, or both, which enhances the visibility and conspicuity of the target but results in low-quality indices of the image as a whole. Second, expressions for the contrast, visibility, and conspicuity of a target available in the literature are not symmetric with respect to zero and some of them are singular functions. We claim that an image quality index describing the target visibility should satisfy, at least, the following requirements. First, the notion of the threshold local contrast, still (or no longer) resolved by human eyes, must be involved in the definition of visibility. Second, the visibility index should not be a singular function. Third, visibility is a notion strongly related to the neuronal response of the human brain and, thus, should be in the form of an activation function. Forth, the target visibility index should be of the probability character. Fifth, the visibility index should be target-oriented. In this paper, we propose an expression defining the target visibility index which satisfies all five requirements listed above and develop a technique for its measurement based on the measurements of brightness profile along a line and calculation of the local contrast of the target. The measurements of the target visibility index are illustrated for the partial (visible and infrared) images and for images fused by algorithms of different target-oriented image fusion methods.