Clutter Metrics Research Articles

Evaluation of human detection performance using target structure similarity clutter metrics

One of the most challenging research topics in the field of target detection in electro-optical images is the relationship between image content and human detection performance. We propose two target structure similarity (TSSIM) metrics to describe global image clutter. Via a simple mathematical formula, the TSSIM clutter metrics and their combinations with different local target-to-background contrasts, which are loosely referred to as signal-to-clutter ratios (SCRs), are used to predict human detection performance. Other clutter measures and their combinations with the contrast ones are also considered for comparison. The degrees of correlation between target detection probabilities and various clutter metrics, as well as the SCR models, are presented. Experiment results show that the TSSIM measures are more suitable for quantifying electro-optical clutter than other clutter measures compared. And the SCR metric, which combines the root-sum-of-squares local contrast and a TSSIM clutter metric, generates the best predictive capabilities for the experimental detection probabilities.

Effects of 'local' clutter on human target detection

In theory, properties of clutter can be defined globally or locally. However, in the literature, the distinction between local and global clutter is arbitrary, where the standard approach of setting the local domain to twice the expected target size, in applying local clutter metrics, is adopted without any justification. This paper addresses this problem and considers the implications for the application of clutter metrics. It was found that the size of the local clutter region around a target has a strong effect on the probability of detection of that target and that this is affected by regions much larger than twice the target size. It was also discovered that this effect was much stronger for targets subtending less than 0.8 degrees of visual angle than for larger targets. In the case of the former, the fall-off in human visual performance with clutter region size was approximately quadratic, compared to a slight linear fall-off for larger targets. A simple model is presented explaining these phenomena, indicating that the auto-covariance function characterising the clutter is the main determinant of the size of the region of local clutter.

New metrics for clutter affecting human target acquisition

Clutter plays a very important role in the area of human-in-the-loop target acquisition. A great deal of interest has recently been shown in defining clutter for electro-optical images. In spite of so many definitions available, no single clutter definition has been agreed on by the electro-optical community as being the best. We modify the structure similarity index, which is extensively used for image quality assessment, to be clutter metrics and call them target structure similarity measures (TSSIMs). Two different methods for calculating the TSSIM metrics are suggested. The TSSIMs of many images in Search/spl I.bar/2 dataset have been analyzed. Experimental results show the proposed metrics correlate very well to the detection probability of the observers.

Image-based contrast-to-clutter modeling of detection

Abstract : Using image-based metrics, contrast-to-clutter modeling is applied to the Search-2 visible image set and perception experiment data. To calculate the contrast metric, a new image is generated from the original image by replacing the target with an expected using the local background surrounding the target and the natural horizontal correlation present in most surface-to-surface scenes. The contrast metric is obtained from the difference of this new image and the original image. Via a simple mathematical formula, the ratio of the contrast measure to a clutter metric is used to predict performance.

Evaluation of human detection performance of targets embedded in natural and enhanced infrared images using image metrics

The evaluation of human detection performance of targets embedded in natural iR images is uniquely extended to deal with enhanced images. Various image metrics designed to describe global image clutter and local target-to-background contrast are presented. Agreement between experimental results, as obtained for both natural and enhanced iR images and image metrics, is evaluated by correlating the metric values with target detection results obtained from two psychophysical experiments, which were designed to record and measure human target detection performance and image quality judgments. Positive correlation values are generally obtained between the metric values and both the target detection (rate and probability) performance and the image quality scale values. More specifically, the local contrast metric is found to be the most suitable to the data recorded from the quality judgments experiment, while the global clutter metric is found to be the most suitable to the data recorded from the detection performance experiments. Further analysis of the relationship between the image metrics and the experimental results yield empirical classification thresholds that can be used to evaluate and predict human detection performance in similar cases. A method to evaluate the classification efficiency, is introduced and used to reconfirm the metrics appropriateness to describe detection performance and quality judgments of targets embedded in both natural and enhanced images. © 2000 Society ofPhoto-Optical Instrumentation Engineers. [S0091-3286(00)00202-6]

Influence of sensor motion on infrared target acquisition

The effect of linear image motion and high frequency vibrations on human target acquisition performance is considered. Two clutter metrics, one local and the other global, are combined to form a single signal-to-clutter ratio (SCR) metric. The SCR is used as a parameter in the model for actual target acquisition results. Two experiments involving human observers are considered: (1) A static experiment is performed with spatial filters whose frequency characteristics are those appropriate for the image degradation obtained from a moving sensor, and (2) a dynamic experiment is executed which imitates the operation of a scanning camera with a constant velocity as in airborne imagers. It was found in both experiments that image motion increases the detection time of a target by the observer. As the complexity of the original images increases, detection time was found to increase more rapidly as a function of blur radius in the first experiment and velocity in the second experiment. However, our blurred images in the first experiment, representative of the degradation expected from motion, actually enables more observers to eventually acquire the target despite the longer acquisition time.

Wavelet transforms of cluttered images and their application to computing the probability of detection

Wavelet transforms are currently being used for a number of applications such as cue feature and noise extraction from images and acoustic signals. This work describes and applies an algorithm that uses wavelets for finding the clutter in infrared (IR) and visual images. Once the clutter is found, the probability of detection (Pd) is calculated. Tidhar’s and Rotman’s probability of edge metric (POE) is extended to encompass the wavelet methodology for multiscale clutter metrics in IR images.

Target acquisition and false alarms in clutter

Recent studies on low-observable target acquisition relate clutter to traditional true- and false-detection performance parameters. This paper presents modified definitions of the target signature and clutter in the image plane and their relation to alternative metrics. Measured detection results for experienced observers in a controlled experiment strongly correlate with target signal-to-clutter ratios calculated for the various images. Empirical false-alarm rates were quite constant as a function of time during the typical observer search process; a sudden decrease to zero occurred at a typical "saturation time" for the standard observer. The rate of false alarms and the saturation time were invariant to the particular signal and clutter signature. The algorithms derived on the basis of the theory of signal detection suggest an effective decrease of the threshold signal with increased observation period and attribute the detection-probability behavior with time to this factor alone. Although the behavior with time differed from that predicted from random-glimpse statistics, an algorithm using that approach is also presented and serves as a numerical approximation. The signal and clutter metrics may be applied as tools to compare and rank target detectability and background complexity in various images.

Modeling human search and target acquisition performance: fixation-point analysis

There has been an increase in attempts to quantitatively measure clutter in the background of visual and infrared scenes; this parameter is crucial for determining the probability of detection of military targets in actual battlefield conditions. We analyze several clutter metrics by comparing the location of fixation points with the spatial distribution of the areas of high clutter. We determine that a strong correlation does exist.

Modeling human search and target acquisition performance: IV. detection probability in the cluttered environment

To model the target acquisition capability of an electro-optic system operated by a human, one should take into account how the clutter in the scene affects the human performance. We consider one global and two local clutter metrics; we compare actual target acquisition results to those predicted by our metrics.

Clutter metrics for target detection systems

The problem of calculating an effective contrast for diffused targets in visual scenery is addressed. The effect of local variations of intensity in the vicinity of the target is modeled by considering the distribution of the intensity of the edges. The effect of global clutter is discussed from the vantage point of selecting the best model for background scenery. Generalized contrast measures are described and applied to real scenes. >