Viewers are recommended to maintain comfortable viewing conditions by taking the following actions :
– Adjust light, sound and air condition of the viewing space.
– Adjust brightness and colour of the display.
3 Viewer guidelines
3.1 Symptoms caused by 3D viewing on viewers
Viewers are recommended to stop viewing 3DTV if they have a headache, pain in eyes, dizziness, nausea, palpitation, unclear vision, unpleasant feeling, optical discomfort or double vision .
Fast movement of objects on the display or excessive accommodation-convergence mismatch/conflict may cause eye fatigue. If viewers experience any discomfort, they should stop watching and stare into a far distance to relax the eyes.
3.2 Stereo blindness and stereo abnormality
Viewers with stereo blindness or stereo abnormality may suffer optical discomfort with double vision. They are recommended to avoid watching 3D images.
– Not every person can perceive vivid stereoscopic depth from 3D images. About 1% of the population cannot perceive the stereoscopic depth at all. Up to 30% of the population perceive stereoscopic depth by detecting the stimulation for a specific depth only from the protruded (negative parallax) or retreated images (positive parallax) .
– Viewers having a strabismus or astigmatism may have a difficulty in perceiving the stereoscopic depth, and suffer more severe fatigue of eyes with double image than the viewers with normal vision.
– Amblyopia is accompanied by astigmatism in many children, and generally, these children cannot perceive the stereoscopic depth. Viewers with large difference between left and right eye may have a difficulty in deceiving the stereoscopic depth.
3.3 Chronic diseases
Viewers with any chronic disease (epilepsy, cardiac disorder, high/low blood pressure, etc.) are advised to pay special attention to any discomfort experienced when viewing 3D images.
It is recommended that children under 10 are carefully supervised when viewing 3D images because they do not have fully developed optic systems and functions .
– Children have relatively shorter distance between eyes, and they perceive higher binocular disparity of 3D image than adults. Therefore, children perceive higher stereoscopic depth than adults from the same 3D image.
Ageing may deteriorate the stereopsis function.
– It is reported that ageing reduces the optical capability of deceiving the stereoscopic depth. Therefore, middle-aged/old viewers may have a difficulty in perceiving a vivid stereoscopic sense when compared to the younger age .
4 Content guideline
4.1 Setting stereo cameras
When producing stereoscopic image with stereo cameras, it is important that parameters are applied consistently to the left and right cameras. The following considerations are recommended:
– It is important to adjust the optic axes of the stereo cameras. The vertical adjustment error, rotation adjustment error and intersection error must be minimized.
– The basic camera parameters must be set so that there is no error of zoom, focus, iris and colour between right/left images. It is recommended to have left and right cameras synchronized.
– Vertical inconsistency of images caused by inconsistency of optic axes is known to cause a fatigue of eyes.
– The state that the intersection point between optic axes of left and right cameras is not on the centre lines of the stereo cameras is called an intersection error. An intersection error may occur due to optic axes adjustment error even if the stereo cameras are set properly without any vertical adjustment error or rotation adjustment error. If the intersection error is excessive, the cameras cannot express the symmetric stereoscopic sense on the left/right symmetric stage, but express in appropriate stereoscopic sense.
– Excessive stereoscopic sense continued unnecessarily or sudden change of stereoscopic sense in the course of production of content may easily cause a fatigue of eyes.
Types of error of optic axis adjustment error
4.2 Capturing stereoscopic images
Following considerations are recommended in capturing 3D stereoscopic images.
− Sudden changes in depth can cause eye fatigue. To minimize this, images should be produced so no sudden changes of disparity occur in a short period of time. Smooth camera operations when zooming or panning also reduces the risk of eye fatigue.
− It is also recommended that care should be taken during editing to prevent severe changes of disparity between shots or scenes.
− When taking close-up images with toed-in cameras (where the optical axes of the two cameras intersects), keystone distortion should be minimized and compensating work should be carried out on post-production if the images are not comfortable to watch.
− When using zoom lenses, the cardboard and puppet theatre effect should be minimized by adjusting the space between cameras.
– To avoid edge violations, any object near either side of the image should not be projected out of the screen (negative parallax) if at all possible.
– If the depth of an object perceived through binocular screen disparity is too small when compared to the known depth of the object, a cardboard effect may occur. In this case, this effect should be minimized with careful camera work .
– A puppet theatre effect is made when there is a gap between the stereoscopic image displayed and the size of the object perceived in the real world. Viewers are likely to perceive an object as too small compared to the background. This effect often occurs on smaller displays .
– A disparity between left and right images should be consistent if the image is produced properly. If a scene is taken with toed-in cameras, the image produces a keystone distortion in the echelon formation with inappropriate vertical parallax. The vertical parallax increases as the space between cross stereo cameras increases and the focal distance of the lens reduces .
– Toed-in cameras can create inappropriate horizontal parallax, which in turn, causes a distortion of stereoscopic depth. Viewers may perceive that an object is at a different distance when at the side of the display than when in the centre of the display .
4.3 Captions and Graphics
Captions must be displayed in front of objects to prevent violations that would make the caption appear to be inside an object. If an object has a high negative parallax (is projected a long way out of the screen), then placing captions in front of the object may make it uncomfortable to watch due to excessive disparity. In this case, it is recommended to change the position of the caption to prevent fatigue of eyes.
4.4 Screen disparity
– For uncrossed disparity, the screen disparity between left/right images should not exceed the average gap between eyes.
– For crossed disparity, excessive screen disparity between left/right images may cause a fatigue of eyes.
– Continued viewing of excessive binocular disparity causes a fatigue of eyes. Therefore, it is recommended not to present excessive crossed disparity for long periods of time.
5 Display guidelines
Image crosstalk caused by the display causes discomfort when viewing stereoscopic images. It is recommended to minimize crosstalk by consideration of the following:
– Crosstalk is a phenomenon that occurs due to inadequate separation of left and right images in the display. It varies depending on the display type and the left/right image separation method. In a polarized display (passive), for example, crosstalk is caused by the optical performance of the polarizing filter and the incomplete adjustment between the polarizing filter and the display pixel elements. In a shutter display (active), the pixel response speed becomes the main factor of the cause of crosstalk.
– Crosstalk in the display is an independent parameter to the content, and can be indicated in an objective value for each display. For example, the crosstalk for the left eye is induced from the brightness of the right image versus that of the left image seen by the left eye, and vice versa.
The crosstalk experienced subjectively by viewers may be caused by the content, as well as the display crosstalk. Therefore, it is recommended to consider the following when producing content:
– Subjective crosstalk experienced increases if there is a large contrast or binocular disparity between left and right images at the same position of the display .
5.2 Display refresh rate
In order to prevent flickering, the refresh rate of left and right images should be 60 Hz or higher. Therefore, the total refresh rate of a time-division 3D display should be 120 Hz or higher.
5.3 3D glasses
In addition to the display crosstalk, crosstalk may occur due to the glasses. Therefore, it is recommended to consider the following to minimize crosstalk:
– For polarized glasses, leakage of light caused by optical performance of the polarizing filter generates crosstalk.
– For active shutter glasses, crosstalk is generated by light penetration due to the optical performance of the glasses when the liquid crystal is closed.
– Active shutter glasses acquire information sent from the display for synchronization between left and right images. In order to minimize crosstalk and the signals and protocols for communication must be robust and immune from external interference, and interruption.
 ISO IWA3, Image safety-Reducing the incidence of undesirable biomedical effects caused by visual image sequences, 2005.
 SUGAWARA, M., MITIANI, K., KANAZAWA, F.M., OKANO, F. & NISHIDA, Y., “Future prospects of HDTV – Technical trends toward 1080p”, Paper presented IBC2005 conference, 2005.
 VAN EE, R., “Correlation between stereoanomaly and perceived depth when disparity and motion interact in binocular matching”, Perception, 32, pp. 67-84, 2003.
 RICHARDS, W., “Stereopsis and stereoblindness. Experimental Brain Research”, 10, pp. 380-388, 1970.
 RICHARDS, W., “Anomalous stereoscopic depth perception”, Journal of the Optical Society of America, 61, pp. 410-419, 1971.
 BELL, B., WOLF, E., & BERNHOLZ, C.D., “Depth perception as a function of age”, Aging and Human Development, 3, pp. 77-81, 1972.
 WOODS, A.J., “Understanding crosstalk in stereoscopic displays”, Keynote presentation in 3DSA conference, Tokyo, Japan, pp. 19-21, 2010.
 Lydia M.J. MEETSTERS, Wijnand A. IJSSELSTEIJN, and Pieter J.H. SEUNTIENS, “A Survey of Perceptual Evaluations and Requirements of Three-Dimensional TV”, IEEE Transactions on circuits and systems for video technology, Vol. 14, No. 3, pp. 381-391, 2004.
 Andrew WOODS, “Image Distortions in Stereoscopic Video Systems”, Stereoscopic Displays and Applications IV, Proceedings of the SPIE Vol. 1915, San Jose, CA, 1993.
Examples of safety guidelines Italian Health Ministry Circular Letters
The Italian Health Ministry issued a Circular Letter dated March 17, 2010, addressed to all the Regional Health Agencies, the Police, the associations of cinema operators and for information to the Ministry for Economy Development, which is responsible for telecommunications including broadcasting. The Italian Ministry action was prompted by a request of the Italian Consumer Protection Authority, and it was based on advice provided by the Ministry’s High Advisory Council on Health. Some further clarifications were issued in a subsequent circular letter dated 6 August 2010.
The circulars are available (in Italian) on the website of the Ministry, and a translation is provided below.
The Ministry circular letter of March 17, 2010 states that:
– the scientific literature does not seem to provide firm proof that viewing stereoscopic programming would force human eyes and brain to perform unnatural processing of visual information; consequently there are no clinical indications at this moment against the use of 3D spectacles during cinema screenings, on condition that such screenings are limited in duration (the advice of the High Advisory Council on Health on this point was more detailed: it is suggested that viewing should be limited in time, and that it should contain intermissions proportionate to the total programme duration);
– some functional problems may arise in young viewers due to the use of 3D spectacles to view cinema presentations, because binocular vision may not yet be present or well established in young viewers, or because they may be cross-eyed or amblyopic, or because they may be going through a period of visus rehabilitation; however those problems should cause no irreversible damage or pathologies;
– consequently, the public that attends stereoscopic screenings should be informed that children under six years should not use 3D spectacles and that even adults should not use them for a duration that exceeds a single screening session.
A further circular letter of the Ministry, dated 6 August 2010, stated that in those cases when single-use glasses cannot be considered due to their technology or cost, 3D glasses must be properly disinfected and repackaged before they are provided to the next user since the risk of transmission of bacterial infections tends to increase with the successive use of the same spectacles by different viewers.
Psychophysical studies on three dimensional television systems
Before attempting to implement new broadcast schemes, we must gain a full understanding of the results of psychophysical studies in order to understand the effects to which the viewer is subjected and the performance that is required of the main equipment in these systems.
There are a number of issues to be studied before we can fully understand the effects of viewing three-dimensional images on human perception and visual functions. For the success of three‑dimensional television broadcasting, all parties concerned, including broadcasters, producers, manufacturers, and regulators, should be well informed of the effects.
The psychophysical aspects of viewing stereoscopic images have been extensively studied. This Annex provides some key study items and the study results on the psychophysical aspects of stereoscopic television systems. It also describes the spatial distortion prediction system for 3DTV that calculates the spatial distortion of a reproduced stereoscopic image and predicts unnatural size distortion, excessive binocular parallax, and excessive parallax distribution on the basis of the shooting, display, and viewing conditions.
1 Key items for psychophysical studies
The following sections describe the key items for which further study is encouraged:
1.1 Naturalness and unnaturalness of images
1) Theoretical analysis of spatial reproduction characteristics of images taken by 3D cameras
It is of fundamental importance to understand precisely how a real space is converted into a stereoscopic image space by a camera. In particular, the reproducibility of a stereoscopic image space should be analysed in terms of different settings of the lens axes of 3D cameras.
2) Size distortion
The reproduction magnification ratio of an object at the shooting distance (the perceived size) varies with the imaging and display conditions. The resulting distortion in size may make an object be perceived as unnaturally small; this is called the “puppet theatre” effect.
3) Depth distortion
The imaging and display conditions may reduce the reproduction magnification ratio of the depth direction and distort the perception of objects with visually imperceptible thicknesses. This is called the “cardboard” effect.
Section 2 describes the study results on the naturalness and unnaturalness of stereoscopic images.
1.2 Viewing comfort and discomfort
1) Differences in size, verticality, inclination and brightness, and cross-talk
Viewers may not feel comfortable viewing left and right images that have size, verticality, inclination, and brightness differences. Cross talk between the left and right images may also have an impact on viewing comfort.
2) Psychological factors and the parallax distribution
The fundamental relationship between psychological effects brought about by 3D images and factors related to fatigue should be studied. In particular, “ease of viewing” and “sense of presence” may be key psychological factors. Attention should be paid to the distribution of parallaxes in the stereoscopic images. From the correlations between psychological factors and the parallax distribution, we can grasp the essential characteristics of stereoscopic images, e.g. the sense of presence they convey and their ease of viewing (visual discomfort).
3) Superimpositions within 3D images
With regard to superimpositions in a two-dimensional image, we only have to think about exactly where to display it on the screen. In the case of a stereoscopic image, however, we also need to pay attention to the depth of the superimposition. If we could find a preferred position for superimposition for stereoscopic images, we will be able to use it for actual programme production.
4) Change in parallax distribution during scene changes
The parallax distribution of stereoscopic images is discontinuous during scene-change frames, where the scene depth and perceived convergence distance change. We need to evaluate how these changes affect the visual discomfort experienced during viewing of stereoscopic images.
Section 3 describes the study results on the viewing comfort and discomfort of stereoscopic images.
1.3 Visual fatigue caused by parallax 3DTV viewing
Visual fatigue caused by viewing stereoscopic motion images is a particular safety concern. Viewer’s repeated adaptation to the discrepancy between eye convergence and accommodation causes a decline of their visual functions and results in visual fatigue.
Section 4 describes the study results on the visual fatigue caused by viewing stereoscopic images.
1.4 Individual differences in the stereopsis function
Visual functions vary greatly from person to person; so it is essential to understand that there are individual differences before 3D broadcasts begin. For instance, there are limits to the binocular parallax of left and right images that a person can fuse into one image; when the parallax exceeds these limits, a double image is perceived. In this situation, depth perception collapses and viewing becomes extremely uncomfortable. For this reason, it is necessary to know the range of binocular parallax over which two images can be fused into one. However, individual differences are vast and will necessitate a study of the stereopsis function of many people.
1.5 Effect on young people
We must bear in mind that young people’s sense of sight changes as they mature. Viewing of stereoscopic images possibly affects their visual functions in ways different from adults. It may be advisable that young children be cautioned about viewing stereoscopic images for extended periods of time.
2 Naturalness and unnaturalness of stereoscopic images − Geometrical analysis of spaces reproduced by stereoscopic images
2.1 Theoretical analysis of reproduced spaces
A basic requirement for the design of stereoscopic systems is an understanding of the transformation from real space (the space in which an actual object exists) to reproduced stereoscopic image space (the representation of this space in a stereoscopic image). In this section, we analyse the distortion of reproduced stereoscopic image space on the basis of image shooting and display system parameters .
2.1.1 Model of shooting/display systems
The configurations of the image shooting and display systems analysed here conform to the parameters shown in Fig. 4. The details of these parameters are shown in Table 1.
Shooting and display systems can typically be configured in two different ways depending on how the optical axes are arranged.
Parallel configurations (where the two cameras of the stereo camera are aligned parallel to each other) are characterized such that objects at infinity are displayed at infinity by maintaining a constant horizontal separation of Hc between the left and right images when they are displayed (see Fig. 4). As a special case, when the separation dc between the cameras and the horizontal offset Hc between the left and right images are equal to the separation de between the viewer’s pupils, and the lens angle b is equal to the angle of view of the display screen d, the real space is in theory reflected without distortion in the reproduced stereoscopic image space . However, it is not always possible to satisfy this condition in broadcasting where a wide variety of different subjects are liable to be viewed under widely varying conditions.
Another optical axis configuration is the so-called toed-in configuration wherein the optical axes of the two cameras intersect (see Fig. 5). This configuration is characterized such that an object situated at the intersection of the optical axes appears at the depth position of the screen on which its stereoscopic image is displayed. It is also relatively easy to present a sense of depth for objects in the space in front of and behind the object at the intersection of the optical axes. By virtue of these characteristics, this method appears to be used in most stereoscopic programmes.
Parameters in shooting and display models
Position of a stereoscopic object
Angles of view of lens
Camera convergence angle
Convergence angle of eye
Horizontal gap between L and R images
Width of screen
Width of virtual screen at the viewing distance in the shooting model
Distance from the centre of the virtual screen at the viewing distance in the shooting model (see Fig. 4)
If an object’s depth position in real space (the environment where images are shot) and stereoscopic image space (the reproduced environment) are Lb and Ld, respectively, then the relationship between these values obeys the following formula using the parameters of Table 1 and the geometrical relationship of the system configuration shown in Fig. 4.
In a parallel configuration, we can set Lc→∞ and Hc = de. In a toed-in configuration, we can set Hc = 0.
Table 2 shows the results of using equation (1) to investigate how the depth position Ld in the reproduced image and the actual depth position (the original camera-to-object distance) Lb are expressed in systems with parallel and toed-in configurations.
In Table 2, no consideration is given to the keystone distortion of the image shape that occurs in toed-in configurations. In other words, this Table shows the characteristics at the centre of the image where keystone distortion has little effect.
In the parallel configuration, Lb and Ld obey a proportional relationship regardless of the parameter settings. On the other hand, in the toed-in configuration, Lb and Ld are equal only for a certain specific combination of parameters (Lc = a1·a2·Ls), but otherwise have a non-linear relationship. The graph of Table 2 indicates that different characteristics are exhibited depending on the sizes of Lc and a1·a2·Ls.
Distances in real space and in reproduced stereoscopic image space