Moving-Picture Response Time and Perceived Motion Blur on LCD Panels

It is only in the last few years that the problem of motion blur became fully understood, but real progress requires an objective measurement technique that predicts what viewers actually see.

by Jun Someya

THE use of liquid-crystal displays (LCDs) in monitors, televisions, and other display devices has rapidly propagated in recent years. As this rapid, worldwide popularization of LCDs progresses, the motion blur of moving images on an LCD screen is becoming a significant problem.

Motion blur on LCDs is the phenomenon in which the contour of a moving image appears blurred to viewers, and it has two different causes. The first cause is the response time of the liquid crystal (Fig. 1). The 3-D diagram on the left-hand side of the figure illustrates typical liquid-crystal response characteristics at different gray levels. The x-axis shows the initial gray-level values of pixels, the y-axis shows gray-level values after they are changed, and the z-axis shows the response time. The longer the response time, the more prominent the motion blur becomes.

The diagrams on the right-hand side of Fig. 1 show the change in image data and luminance versus time. In the example illustrated, it took 40 msec to change the luminance when the gray level changed from a value of 32 to a value of 244. On a conventional LCD TV, images are updated every 16.7 msec, which is one frame period. Therefore, the response time of LCDs must be significantly shorter than 16.7 msec if motion blur from this cause is not to be an issue. For the panel defined in Fig. 1, motion blur would be obvious. Research and development to improve the quality of liquid-crystal materials and to shorten their response time by implementing signal processing has been ongoing, and some of the advancements have been incorporated into current production panels.

 

Figs-1_tif

Fig. 1: The first of the two major causes of motion blur is the response time of the liquid crystal. When a pixel with a particular gray level specified on the x-axis of the 3-D plot receives image data instructing it to change to another gray level specified on the y-axis for the TFT-LCD described in this figure, it does so with the response time indicated on the z-axis. The response time (right, bottom) is the time it takes for the pixel to reach its intended luminance after the image data has been changed.

 

But that is not the whole story. It is also known that motion blur occurs even if the response time of the liquid crystal is reduced to 0 msec. This blur is caused by the sample-and-hold method by which images are displayed on LCDs, and this is the second cause of motion blur. Let us look at a situation in which a white object moves from left to right across an LCD screen on a black background, where the horizontal axis represents the horizontal position on the LCD's screen and the vertical axis is time (Fig. 2). The solid line on the diagram on the left indicates the relationship between the central position of the white object and the time when the white object is displayed at the same position for a period of one frame. This white object moves by one frame displacement, as in a single-frame advance.

When viewed, the human eye smoothly traces the white object as if there was actual movement, and the point of view also moves smoothly as shown by the broken line. Figure 2 (right) presents the situation from a different point of view, i.e., with the horizontal axis now representing the image's horizontal position on the retina. What is actually happening is that a moving edge displayed on a hold-type display vibrates from left to right on the retina, and this vibration is perceived as a motion blur. How can this second type of motion blur be reduced?

Reducing Sample-and-Hold Blur

Because explanations of motion blur in LCDs based only on the response time of the liquid crystal were clearly inadequate, several companies gathered in 2001 and began discussing methods for evaluating motion blur on LCDs. As a result, a metric called Moving Picture Response Time (MPRT) was defined to quantify the degree of motion blur on LCDs. Today, serious discussions are under way within the Video Electronics Standards Association (VESA) to standardize the definition and measurement of MPRT. Measurement methods for motion artifacts are currently under discussion, and the concept of MPRT has been included in the update document Flat Panel Display Measurement (FPDM), Ver. 2 as one of the measurement methods for motion blur on LCDs.

Measuring MPRT

Definition of MPRT. MPRT involves a technique that quantifies the degree of blur a viewer senses when looking at a moving edge displayed on an LCD. As is implied, MPRT is the duration of time from the point at which a blur arises in the contour of an image in motion until the time it disappears, expressed in milliseconds.

An easier way of understanding the degree of motion blur is to use the width of blur on the displayed image. But for the sake of taking measurements that characterize a display, this intuitive approach presents some difficulties. Specifically, the width of measured blur on the panel depends on the speed of the moving image, the visual range, and the hold time. Therefore, a unit of time is used so that values measured under different conditions may be directly compared. It is then possible to determine the width of blur on the panel if a visual range, speed, and other conditions are provided in addition to the MPRT value obtained by measurement.

 

Figs-2_tif

Fig. 2: Motion blur can also occur on an ideal hold-type display having a response time of 0 msec. In the frame of reference of the viewer's retina, the image vibrates left and right, which is seen as motion blur.

 

Measurement Apparatus. To measure MPRT, a pursuit camera system is used that tracks and takes a picture of the edge of the moving image to be evaluated, moving from left to right by rotating a mirror in front of a camera (Fig. 3). In this method, the critical factor is the accuracy in tracking the movement of the image to be evaluated. Some investigators have reported using systems that utilize a video camera, but a still camera can be used if it has the ability to track the image to be evaluated.

Since, as has been shown, response time depends on the initial and final gray-level values, the determination of a single metric intended to characterize a display must make use of measurements at different gray levels. The measurements required to determine MPRT were made on 42 images that combine seven luminance levels from Y0 to Y6 (Fig. 4). The pursuit camera system captured those images while they were scrolled from left to right.

The luminance levels Y0 to Y6 are based on a range between the lowest lightness (L0) and the highest lightness (L6 = 100) values on the lightness axis L* of CIE 1967, as determined by

L0 = 903.3 x Y0/Y6,

where

Y0/Y6 ≤ 0.008856,

Yn = Y6 eq_1

n ∈ {0, 1, 2, 3, 4, 5, 6}.

Because the luminance intervals from Y0 through Y6 obtained by this method are subjectively equidistant, we can obtain luminance values that are subjectively juxtaposed at even intervals even when the measurement is made on panels having different gamma characteristics. This allows us to take measurements that have no bias when the luminances are combined.

Calculating MPRT. MPRT is the average value of Extended Blurred Edge Time (EBET) measurements determined from the edges of the 42 captured images. First, by using the thresholds of 10 and 90% of the relative luminance for these response characteristics, the transition time was determined, and then this transition time was divided by the width between the two thresholds (0.8) to obtain the EBET (Fig. 5).

 

Figs-3_tif

Fig. 3: This pursuit camera system for MPRT measurement tracks the edge of the image moving across the screen by using a rotating mirror.

 

Figs-4_tif

Fig. 4: Since response time depends on the initial and final gray levels, the determination of a single metric intended to characterize a display must make use of measurements at different gray levels. The measurements to determine MPRT were made on 42 images that combine seven luminance levels from Y0 to Y6.

 

If an LCD has a response time of 0 msec, the slope of this response characteristic curve will become a straight line, and the EBET will be equal to the hold time. In addition, if the frame frequency is increased or if the mode is changed to impulse display to shorten the hold time, then the EBET will decrease, and if the response time of liquid crystal becomes longer, the EBET will also become longer. Since the EBET takes the hold time of an ideal hold display as its base value, it is an index that allows viewers to easily envisage the response characteristics of liquid crystal.

EBETs were determined for each of the 42 evaluation images, and those values were averaged to produce an MPRT as follows:

eq__2

eq_3

Subjective Evaluation and MPRT

MPRT is a valuable metric only if the measured result correlates well with the degree of motion blur that viewers perceive. We have examined this correlation and found it to be extremely strong. The fact that MPRT quantifies the degree of perceived motion blur indicates that the MPRT can be used as an index for the moving-picture display performance of LCDs.

In this study, we examined the degree of perceived motion blur by displaying a series of images on both a CRT and LCD monitor for subjective evaluation (Fig. 6). The LCD monitor was placed on top of the CRT monitor, and a moving-picture pattern was supplied to both monitors. A pattern always free of blur was used for the edge section of the images to be displayed on the LCD. When this image was scrolled from left to right, motion blur was perceived because of the slow response time of the liquid crystal and the affects of the sample-and-hold method.

In addition, the test subject was allowed to use the keyboard to change the blur width of the edge section of the image to be supplied to the CRT. Thus, the subject, while looking at the image that is being scrolled, adjusts the degree of blur in the pattern displayed on the CRT so that the blur width of the edge displayed on the CRT will look the same as that on the LCD. This adjusted blur width was recorded, taken as the degree of motion blur that the subject perceived, and the values were compared with those for the EBET.

 

Figs-5_tif

Fig. 5: The MPRT is the average of the EBET measurements. The EBET is measured by using thresholds of 10 and 90% of the relative luminance for the response characteristics to determine the transition time. That transition time is then divided by the width between the two thresholds to obtain the EBET.

 

Figs-6_tif Mitsubishi Electric Corp.

Fig. 6: The degree of perceived motion blur was determined by presenting a series of images on both a CRT and LCD monitor for subjective evaluation.

 

A comparison of the blur width (in pixels) obtained from the subjective evaluation experiment and the EBET measurements collected from the MPRT measurement shows excellent correlation (Fig. 7). The ordinate corresponds to the width of perceived motion blur using a number of pixels, and the abscissa represents the EBET (msec). By using thresholds from 10 to 90% (or from 90 to 10%), the width of perceived motion blur was determined in the same manner as the EBET was determined from the shape of the luminance of the edge section displayed on the CRT, and the width was multiplied by a factor of 1.25. To measure the EBET, an Otsuka Electronics MPRT-1000 unit was used.

The subjective evaluation experiment, using the 42 images used in the MPRT measurement, produced a value of 0.938 for the coefficient of correlation between the perceived motion blur width and the EBET. A more in-depth analysis resulted in a slightly different trend between the falling edge (from Y6 to Y0, for example) and the rising edge (such as Y0 to Y6). The correlation coefficient for the former was 0.981 and that for the latter was 0.779. The detailed data supports a strong correlation between the results of MPRT measurement and subjective evaluation.

Conclusion

Values obtained from MPRT measurements, which is on its way to standardization, have a strong correlation with the degree of perceived motion blur, and this method is suitable for providing an index to evaluate motion-picture blur on LCDs. Today, VESA is working on document FPDM3 as the standardization of an evaluation technique for motion artifacts. Because there is a strong demand for earlier completion of standardization from manufacturers of LCDs, set makers, as well as others, the concept of MPRT has been published in an updated FPDM2 document as a preliminary publication for FPDM3. •

 

Figs-7_tif

Fig. 7: A comparison between the blur width (in pixels) perceived by viewers and EBET measurements shows excellent correlation.

 


Jun Someya is Manger of the Imaging I/O Technology Department of the Advanced Technology R&D Center at Mitsubishi Electric Corp., 1 Zusho Baba, Nagaokakyo, Kyoto 617-8550, Japan; telephone +81-75-958-3028, fax +81-75-953-5911, e-mail: someya@img.kyo.melco.co.jp.