Embracing Raycasting for Virtual Reality

Abstract

This paper proposes an acceleration method for direct volume rendering (DVR). Our approach works like a wrapper or braces surrounding raycasting implementations and requires very few changes to the existing code. Visualization systems can significantly improve their rendering performance in virtual reality setups and make DVR feasible in these environments. The first step-the opening brace-modifies the initial ray construction by adaptively reducing the ray density, hence feeding fewer rays to the raycaster. The second brace step is a composting computation after the ray traversal that re-samples the raycasting results across the screen to reconstruct the final image. The rendering resolution is adapted during run-time to the specifications of the VR hardware and the performance of the renderer to guarantee stable and high refresh rates necessary to avoid severe cyber-sickness symptoms. The presented method utilizes gaze-dependent resolution levels tailored towards the human visual system (HVS) and hardware characteristics found in state-of-the-art head-mounted displays (HMDs). The resolution, and therefore the number of processed fragments during ray traversal, is reduced in the peripheral vision, delivering unnoticeable losses in image quality while providing a significant gain in rendering performance.