Hologram Upscaling for Viewing Angle Expansion Using Light Field Extrapolation with Object Detection Algorithm

초록

As demand for high-resolution holographic displays in augmented and virtual reality (AR/VR) increases, the limitations of traditional computer-generated holography (CGH) upscaling methods, including bicubic interpolation and deep learning-based techniques, become apparent. These methods predominantly estimate additional pixels without considering the reduction of pixel pitch, inherently constraining their capacity to effectively expand the viewing angle. Our study introduces a novel approach for viewing angle expansion through light field (LF) extrapolation by applying an object detection algorithm. This process starts by analyzing the object position and depth information of each LF view extracted from CGH patterns with the object detection algorithm. The use of these data allows us to extrapolate LF views beyond their initial viewing angle limit. Subsequently, these expanded LF views are resynthesized into the CGH format to expand the viewing angle. With our approach, the viewing anglewas successfully doubled from an initial 3.54 degrees to 7.09 degrees by upscaling a 2K 7.2 μm CGH to a 4K 3.6 μm CGH, which was verified with both numerical simulation and optical experiments.

키워드