A Stereo Matching Method for Three-Dimensional Eye Localization of Autostereoscopic Display
Bangpeng Xiao,Shenyuan Ye,Xicai Li,Min Li,Lingyu Zhang,Yuanqing Wang
DOI: https://doi.org/10.1007/978-3-030-87361-5_3
2021-01-01
Abstract:In order to reduce crosstalk and support dynamic viewing, the mainstream autostereoscopic display technology requires high-precision three-dimensional eye localization. Due to the need of displaying high frame rate video images, higher requirements are put forward on the real-time performance of the three-dimensional eye localization algorithm. The three-dimensional measurement of the distance of the eye is particularly complicated, and stereo matching usually needs to be done with the help of binocular cameras. Aiming at the problem of low efficiency in conventional stereo matching, this paper improves and optimizes the ZNCC stereo matching algorithm from two aspects. On one hand, the operation logic of the matching algorithm is improved. It uses computer memory to save the intermediate results of the operation and uses a dynamic programming method to reduce the computational complexity of the matching cost function. On the other hand, the scanning strategy is optimized based on the application scenarios of stereoscopic display. Using the characteristics of the application scenarios and the additional constraints information of stereoscopic display, the search area is narrowed to reduce the frequency of the matching cost calculation. Various comparative experiments show that the method has the characteristics of strong real-time performance, strong robustness, and high accuracy, and can achieve great application effects in actual autostereoscopic display systems.