Culling-Based Real-Time Rendering With Accurate Ray Sampling for High-Resolution Light Field 3D Display

Real-time rendering for light field displays (LFDs) enables human-computer three-dimensional (3D) interaction with smooth and realistic experience, which is potentially useful for 3D gaming, virtual reality, or surgical navigation. However, existing rendering methods realize a real-time rendering normally by sacrificing the 3D image resolution. We propose a culling-based real-time rendering (CBR) method with an accurate ray sampling, aimed at rendering the optimal synthetic images in real-time, which reproduce the high resolution 3D images. The synthetic image generation and display processes are first analyzed, identifying the condition of accurate ray sampling for generating the optimal synthetic images. Then the CBR method is proposed, significantly improving the generation speed of the optimal synthetic images under the accurate ray sampling condition by employing triangle culling in geometry shading stage and pixel culling using pre-computed viewpoint mask texture. Experiments show that accurate ray sampling generates the highest 3D image resolution, and the CBR method achieves superior frame rate, typically rendering 39.7 frames per second (fps) for a LFD with 4,225 viewpoints and a model of 16,301 triangle faces.