Perceptibility Parameters for Quantification of Uniform Haze Impairment

Whereas work discussed in the previous sections has emphasized detection thresholds of layered hazes, specifically plumes, other researchers have concentrated their efforts in establishing the change in image appearance required to just notice a difference in image sharpness.

Early work focused on establishing the just noticeable difference between a scene where an object viewed against the same background could just be seen and one where that object could not be identified. This threshold work was carried out in the context of establishing the "threshold" contrast for visual range determination.

More recent work has been directed toward incorporating results of basic psychophysical measurements into models that will predict the change in display modulation transfer function (MTF) required to evoke a just noticeable difference (JND) in display image sharpness. Displays of interest were television-type video displays. One model, the quadratic detection model (QDM), relies on the calculation of the image mean square luminance fluctuation, termed the image modulation depth. Henry (1979) and Henry et al. (1981) have suggested that modulation depth may be appropriate visibility indices because they incorporate all of the information content contained in a scenic vista.

Malm and Pitchford (1989) have suggested using the concept of a just noticeable change (JNC) in the appearance of a landscape feature as a psychophysical variable that relates directly to human perception. A JNC corresponds to the amount of absorbing gas or atmospheric particulate matter required to evoke a noticeable change in the appearance of a particular landscape. The effect of a change in aerosol concentration can then be expressed as the number of JNCs between landscape appearance under current conditions versus the appearance after a change in emissions. Malm and Pitchford (1989) have suggested using the QDM to predict a JNC; however, any psychophysical model relating changes in aerosol concentration to human eye-brain visual thresholds could be used for this purpose. It is emphasized that none of the currently used psychophysical models have been field validated.

More recently, Pitchford and Malm (1994) have proposed the deciview scale, which is based on the fact that all detection threshold models and experiments show that above contrasts of about 0.02, a just noticeable change in contrast, is directly proportional to the initial contrast, AC = LC, where L is a proportionality constant. By assuming the availability of sensitive scenic targets at every distance, it can then be demonstrated that any specific fractional change in extinction coefficient is equally perceptible regardless of baseline visibility conditions. The index is defined so that its scale, which is expressed in deciview (dv), is linear with respect to fractional changes in extinction and is given by, dv = 10 ln(6ex/0.01 km-1), where extinction is expressed in inverse kilometers. A 1-dv change is about a 10% change in extinction.

Application of the Quadratic Detection Model. Typical scenes are made up of features that are quite varied with respect to size, shape, and luminance level. However, some attempts have been made to classify scenic structure into broad categories such as form, line, and texture. Form refers to large shapes seen either against sky or other uniform background, while line is usually associated with appearance of rivers or similar geological features. Texture refers to the periodic contrast associated with sparsely populated trees seen against a uniform background, varied geologic features, or other similar higher frequency scenic structures.

Studies investigating eye fixation and eye motion as observers look at pictures show that pictorial areas with little modulation receive very little attention, while higher modulated scenic features receive more (Boswell, 1975). Since high-contrast edges are most sensitive to changes in atmospheric modulation transfer function and since the discrimination of an atmospheric modulation change in a frequency-specific channel is a minimum when the contrast in that channel is largest, it can be concluded that high-contrast edges are good patterns for predicting the relationship between just noticeable changes in scenic appearance and increases in atmospheric aerosol load.

For many typical scenes, a JNC is equivalent to a change in atmospheric modulation of approximately 0.06. Figure 11 shows a typical JNC surface for an 80% reduction in atmospheric extinction as a function of observer distance and atmospheric extinction, assuming a change in MTF of 0.06 is perceptible. The scattering angle is 15°, ax = Ns where Ns is sky brightness, and the initial contrast, C0, is equal to — 1.0. A typical aerosol mass size distribution with typical chemical properties was assumed.

There are some general features that show up in all JNC surfaces. For any given distance there is a background extinction that is most sensitive to an incremental

Figure 11 Just noticeable change surface plotted as a function of observer distance and atmospheric background extinction. The surface corresponds to a reduction of background extinction of 80%.

change in extinction, and for any given extinction there is observer distance that is most sensitive to extinction change. Secondly, for any given observer distance, the sensitivity of a scene to incremental reductions in atmospheric extinction drastically reduces as background extinction increases; and finally, the distance where the scene is most sensitive to a change in extinction decreases as background extinction increases.

0 0

Post a comment