Premium
21.2: The Design and Optimization of Dual‐Camera based Eye Tracking System for Dynamic Autostereoscopic 3D Display
Author(s) -
Kang Jianghui,
Jia Jia,
Pu Yiying,
Lu Min,
Tan Baolin,
Wu Yu,
Chen Jingyu
Publication year - 2021
Publication title -
sid symposium digest of technical papers
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.351
H-Index - 44
eISSN - 2168-0159
pISSN - 0097-966X
DOI - 10.1002/sdtp.15094
Subject(s) - autostereoscopy , computer science , computer vision , rendering (computer graphics) , stereoscopy , latency (audio) , artificial intelligence , computation , eye tracking , virtual reality , low latency (capital markets) , computer graphics (images) , algorithm , telecommunications , computer network
Eye‐tracking based autostereoscopic 3D display is a promising technology to provide high resolution and low crosstalk stereoscopic images to the user without wearing any glasses. However, current systems fail to provide satisfactory viewing experience due to inherent latency caused by the tracking camera system and algorithm computation time and the low reliability of tracking algorithms, especially when users are in fast motion. High instantaneous crosstalk and intermittent interference patterns will be observed, which significantly degrades the immersive level of the viewing experience. This becomes even worse when the autostereoscopic 3D display is used to provide a simulated reality (SR) experience, in which the user constantly moves the head to explore the virtual reality space from different viewing directions. To solve the problem, this paper proposes a dual‐camera tracking system, focusing on the key design factors to improve its tracking reliability and reduce the user perceived latency. For the latter, it leverages motion prediction and compensate the latency with predictive rendering technique. To measure the performance, this paper defines the Perceived Autostereoscopic 3D Display Motion to Picture (PAMTP) latency to represent the impact of the system latency on the viewing experience. PAMTP latency and the instantaneous crosstalk are evaluated via simulations and experiments, which show our method can reduce the instantaneous crosstalk by nearly 70% and provide highly reliable tracking.