A Premixed Autostereoscopic OptiX-based Volume Rendering

Benjamin Battin, Gilles Valette, Julien Lehuraux, Yannick Remion et Laurent Lucas

Present multi-view display techniques are often based on the use of multi-pass algorithms, one per view usually besides an additional step for image compositing. Generally linearly dependent on the number of views, these approaches introduce a bias by incorrectly considering that depth information is located at the pixel level, while the sub-pixel level would be more suitable to improve the intrinsic quality of the resulting images. In this paper, we introduce a new method for rendering multi-stereo views at interactive frame rates which is innovative on two main aspects: first it is based on an OptiX ray-tracing engine, second it uses a single premixed rendering pass operating at sub-pixel level. This approach has been tested on a specific volume rendering algorithm, natively supporting 3D displays. After a detailed description of our method, we present performance results for different use cases on an 8-view autostereoscopic 3D display. Quantitative aspects of the latency as well as resulting perceived image quality are measured for the proposed method and compared with a conventional a multi-pass GPU approach.

Mots cl├ęs

Real-Time Autostereoscopic Volume Rendering, OptiX Engine, GPU