With the idea that if head-up displays (HUDs) are increasingly common in high-end automobiles, 3D HUDs are a logical next step, a doctoral student and a group of researchers in Germany recently set out to create a system to study crosstalk in 3D HUDs. (Crosstalk in such displays is likely to present significant challenges.) Their goal was not to solve actual crosstalk problems at this point, but to develop an accurate system that would allow for such research to proceed in the future.
The resulting Journal of the SID paper, “Exploring crosstalk perception for stereoscopic 3D head-up displays in a crosstalk simulator,” by Simone Höckh, Annette Frederiksen, Sylvain Renault, Klaus Hopf, Michael Gilowski, and Martin Schell, was recently named the JSID Outstanding Student Paper of the Year for 2015. The paper was selected by JSID ’s associate editors through a voting process. “The selection committee appreciated the versatility and potential of the reported simulation setup, which enables its application to the very specific but very relevant case of stereoscopic head-up displays used in vehicles,” says Herbert De Smet, a Professor at Ghent University and Editor-in-Chief of JSID.
Höckh, Frederiksen, and Gilowski are with the Corporate Sector for Research and Advance Engineering at Robert Bosch GmbH in Stuttgart. Höckh is a Ph.D. student at the Technical University Berlin who anticipates a Dr.
rer. nat. degree in 2016. She also participated in a Ph.D. program at Robert Bosch GmbH from 2013 to 2016, with Gilowski serving as her internal Ph.D. supervisor. She is now employed at Bosch. Schell, who is a professor at Technical University Berlin, is Höckh’s university thesis advisor. He is also with the Fraunhofer Institute for Telecommunications, Heinrich Hertz Institute, in Berlin, as are Renault and Hopf. The research described in the paper was conducted at Bosch.
Head-up displays in automobiles show information, such as the speed and the speed limit, in such a way that it “floats” beyond the windshield in front of the driver without blocking the view. In some cases, HUDs can augment reality by digitally emphasizing road hazards and oncoming cars. Höckh’s research team described an autostereoscopic see-through augmented-reality HUD system that could show several planes or virtual-image distances in such a way as to sort relevant information for the driver. For example, information about the car, such as its speed, could be shown closer to the driver, whereas navigational information, such as arrows for turnoff, could appear farther away. Given the pace of display development for automobiles, “an autostereoscopic 3D HUD seems a logical consequence,” says Höckh.
Such a system does not exist yet commercially, but if it did, note the authors, crosstalk, in which “ghost” images appear to the viewer as a result of the separation of right and left views, would be a major issue. Crosstalk not only creates a lack of fidelity, but eyestrain and headaches. It could also prove a dangerous distraction to the driver.
The researchers set about creating an ambitious system capable of studying crosstalk in a HUD. The hardware consisted of a projector (controlled by a workstation), mirror, horizontal screen, image, and a partially reflecting glass plate (see Fig. 1).
Fig. 1: For the setup of the stereoscopic 3D head-up-display simulator, an observer was positioned on the left of the setup and looked through the glass plate to the right.
One of the biggest challenges of the setup, says Höckh, involved the software. “Usually in 3D visualization software,” authors Hopf and Renault explain, “two perspectives are rendered separately. For a simulation of crosstalk, image sections of both views have to be combined.” For this purpose, the team members from Fraunhofer HHI developed a novel rendering pipeline within Fraunhofer’s 3D-API, Workbench3D, which supported a scalable transfer of image information between the left and right view.
The study included 24 subjects (19 male and 5 female) with an average age of 30.5 years and vision that was either normal or corrected to normal. The test sessions consisted of two parts: the first addressed changes in the participant’s visibility threshold and the second, the variation of the acceptability threshold of crosstalk depending on screen parallax, object type and color, illumination, and contrast. Each participant went through both parts of the study consecutively.
The objects initially appeared without any simulated crosstalk. The supervisor slowly increased the crosstalk value until the participant announced the visibility of crosstalk. Then, researchers asked the participants to adjust the crosstalk value themselves, thereby allowing them to increase and decrease the simulated crosstalk prior to their decision for their individual acceptance threshold. Both visibility and acceptability were recorded.
The researchers found that both crosstalk
visibility and acceptability thresholds depended most strongly on contrasts. The higher the contrast, the lower the thresholds. These and other results, they report, appear to be consistent with existing literature relating to studies performed with conventional 3D displays. In this way, the results indicate the functionality of the system the team created. “However,” they wrote in their article, “further research, especially including a realistic background designed to operate an augmented-reality application, has to be performed.”
Höckh explains that her team’s work was part of an industrial and application-oriented research project
designed to identify user-oriented requirements that such a system would have to fulfill. She adds that the results presented in the paper could also be interesting for other autostereoscopic 3D augmented-reality applications – not just HUDs for vehicles. “As both the hardware and the software are designed to allow for the variation of many parameters associated with the perception of crosstalk, the simulator is a very versatile and potent setup regarding further investigations of crosstalk disturbance in a stereoscopic 3D HUD,” she concludes.
Each year a subcommittee of the Editorial Board of JSID selects one paper for the Outstanding Student Paper of the Year Award, which consists of a plaque and a $1000 prize. The award is sponsored by LG Display. •