Improving the computation efficiency for computer-generated holography

Abstract

Consumer-grade 3D displays are likely to become much more demanding over the next few years as the advancement of fabrication capability in the entire display technology industry.

Computer-generated holography (CGH) based picture generation unit is a promising and viable technology. Unlike other display counterparts where light intensity modulation is being used, CGH can modulate the phase of the incoming light and form an image via interference. This allows CGH to position the wavefront of the display content in mid-air precisely, enabling itself to be a true 3D display.

One of the key bottlenecks that prevent the mass adoption of CGH is the lengthy computation time. This is mainly due to the complex mathematical expression that describes light diffraction. To solve this problem, a non-iterative phase retrial algorithm called one-step phase retrieval (OSPR) was created in 2006. Inspired by the non-iterative nature, we managed to design two dedicated computational systems, one for cost-optimised and the other for high performance, using the hardware description language, and successfully deployed the architectures on field-programmable gate arrays (FPGAs). Furthermore, aside from hardware acceleration, we combine the human visual system and propose a holographic foveated rendering method to increase the calculation speed while maintaining the perceived image quality.

This thesis sets out to investigate and implement the methods to improve the computation efficiency for computer-generated holography. Taken together, these investigations provide important insights into the realisation of a real-time calculation systems with small form factors for the next generation of holographic displays. We present our simulation and experimental results for designed hardware circuits and foveated rendering algorithm throughout this thesis.