Proposal of a quality-index or metric for soft copy display systems: contrast sensitivity study

Abstract

In addition to the inherent qualities of a digital image, the qualities of the monitor and graphics control card as well as the viewing conditions will affect the perceived quality of an image that is displayed on a soft copy display (SD) system. With the implementation of picture archiving and communication systems (PACS), many diagnoses are being made based on images displayed on SD devices, and consequently SD quality may affect the accuracy of diagnosis. Unlike the traditional film-on-lightbox display, optimal SD system parameters are not well defined, and many issues remain unsettled. In this article, the human observer performance, as measured by contrast sensitivity, for several SD devices including an active matrix liquid crystal flat panel monitor is reported. Contrast sensitivities were measured with various display system configurations. Experimental results showed that contrast sensitivity depends on many factors such as the type of monitor, the monitor brightness, and the gamma settings of the graphics card in a complex manner. However, there is a clear correlation between the measured contrast thresholds and the gradient of the display device's luminance response curve. Based on this correlation, it is proposed to use the gradient of luminance response curve as a quality-index or metric for SD devices.

Figure 1

a. The luminance response curves of the Clinton Electronics D13000 AMLCD flat panel monitor measured at 90 and 45 degrees to the surface of the monitor. b. The close-up view of the luminance response curves in Figure 1a. Note the “steps” on the curves. c. The gradient of luminance response curves for the flat panel monitor. Note the spikes and troughs in the gradient of luminance response curves.

Figure 2

The luminance response curve for one of the CRT monitors evaluated in the study. Note the absence of the “step-like” features in Figure 1a and 1b.

Figure 3

a. The contrast threshold and gradient of luminance response curve for the AMLCD flat panel monitor. b. Close-up view of Figure 3a. Note that every time the gradient is less than 0.01, the contrast threshold is larger than the minimum, 1 pixel level.

Figure 4

The contrast-detail curves (a-e) of the Image System M21Max CRT monitor at various background luminance levels and effect of gamma settings.

Figure 5

The effect of changing gamma card settings on the measured contrast threshold.

Figure 6

The contrast threshold plots as a function of the luminance level for a graphics card gamma setting of 1.50 (a) and 2.20 (b) and 3.0 (c). Over-plotted is the gradient of the corresponding luminance response curve.

Figure 7

Contrast threshold versus gradient plot for a CRT monitor (Figure 3) and for an AMLCD monitor (Fig 6). Both a and b show that as the gradient increases, the contrast threshold decreases. Also note that when the value of the gradient is larger than 0.03 almost all the measured contrast thresholds become 1 pixel value, which is the best achievable. In other words, the observer can resolve a low-contrast object with minimum contrast (1 pixel value) in a uniform background.

Figure 8

Effect of room light on the gradient of luminance response curve.

Figure 9

The process from image acquisition to display for a film-screen (hard copy display) system.

Figure 10

The process from image acquisition to image display in a digital system.