Parallax Distribution and Visual Comfort on Stereoscopic Hdtv

Watching stereoscopic HDTV sometimes causes visual discomfort or fatigue. In this paper, we report two results of recent research on the relationship between visual comfort and parallax on stereoscopic HDTV. One is the relationship between visual comfort and parallax distribution in a frame, and the other is the dependence of visual comfort on the magnitude of parallax and motions in some scenes of stereoscopic HDTV programs. Therefore, we measure the magnitude of parallax and motion using the block-matching method, and execute two types of test for evaluating visual comfort. The analysis results indicate that the features of parallax distribution in a frame are strongly related to visual comfort and for greater comfort when viewing stereoscopic images, the upper part of the screen should be located further away from the viewer with less parallax dispersion, and the entire image should be positioned at the back. Further, the magnitude of parallax, motion and these changes affect visual comfort when watching moving scenes, and large parallax causes discomfort even if the motion is small. INTRODUCTION In Japan, digital HDTV has already started, using broadcast satellite. Development work is progressing on cameras, recording equipment and display devices for HDTV-based stereoscopic HDTV, and stereoscopic HDTV programs are presented at special events and at some permanent public facilities. Many stereoscopic HDTV programs have already been produced, including materials shot at the Salt Lake City Winter Olympics and the Sydney 2000 Olympic Games. Compared with conventional 2-D HDTV programs, these stereoscopic programs provide a powerful sense of presence and immersion, but prolonged viewing can cause fatigue and some scenes are uncomfortable to watch . If stereoscopic HDTV is to be widely accepted as a broadcasting method, the images must be comfortable to view and cause minimal visual fatigue. NHK has therefore been combining its production of trial programs with research into the causes of fatigue and what it is that makes scenes comfortable or uncomfortable to watch. The underlying causes of visual fatigue and discomfort are thought to be the following: (1) Geometric distortion in left and right images in the shooting or display system, (2) Crosstalk or differences in the electrical properties of both systems for left and right images, and (3) Contradictions of the convergence and accommodation based on binocular parallax. Recently, we have been studying factor (3), focusing on the dependence of visual comfort and visual fatigue on the image properties, that is, distribution and magnitude of parallax and motion. In this report, we discuss the results of research on two aspects of visual comfort relative to parallax distribution. One is research on the sense of presence and visual comfort relative to parallax distribution for almost still images, and the other is research on the dependence of visual comfort on the amounts of parallax and motion for some programs with moving images. RELATIONSHIP BETWEEN “SENSE OF PRESENCE” AND “VISUAL COMFORT”, AND PARALLAX DISTRIBUTION We conducted a subjective evaluation test using 10 stereoscopic HDTV images, measured the parallax distribution for the images used, and analyzed the correlations between the parallax distribution and the results of subjective evaluation. Evaluated Images As images to be evaluated, we used 10 scenes included in the standard stereoscopic chart in which factors such as the shooting conditions and the positional relationships of the objects are known. Each of the scenes used for evaluating images was 15 seconds long, and did not include any large movement. Evaluation Test We conducted a subjective evaluation test using a total of 20 images, consisting of the 10 scenes mentioned above and an additional 2-D image from each scene. Table 1 shows the outline of experiment 1. The display system used was a stereoscopic HDTV projection system with polarizing glasses and screen size of 120 inches. In the test, first the 2-D image was presented as a reference, then the stereoscopic image was presented, and then the process was repeated twice for each image. In the evaluation of the 2D images, the left-hand image was input to both left and right projectors so that the images could be evaluated while the subjects were still wearing the polarizing glasses. Experimental Results Figure 1 shows the distribution of each of the evaluation images with regard to “sense of presence” and “visual comfort” which are the evaluation terms extracted by the factor analysis. The numbers in the figure 1 refer to the 10 images. The figure clearly shows that, when compared with the 2-D images, the stereoscopic (3-D) images had a strong sense of presence, and that some of the images were comfortable to view but some were not. 21.4 cd/m2 Peak brightness 4.5 m Viewing distance 120-inch / 16:9 Screen size Stereoscopic HDTV projection system with polarizing glasses Display system Twice repetition of 2D-image and 3D-image Presentation 99 ( ages ranging from teens to 60's ) Subjects 10 scenes (10 stereoscopic images and their 2-D images) Evaluated