Single-pass Multi-view Volume Rendering

In this paper, we introduce a new direct volume rendering (DVR) algorithm for multi-view auto-stereoscopic displays. Common multi-view methods perform multi-pass rendering (one pass for each view) and subsequent image compositing and masking for generating multiple views. The rendering time increases therefore linearly with the number of views, but sufficient frame-rates are achieved by sub-resolution rendering, at the expense of degraded image quality. To overcome these disadvantages for DVR, our algorithm calculates multiple views directly on a per-fragment basis on the GPU in a single rendering pass, including sub-pixel wavelength selective views for high-quality auto-stereo display systems. Moreover, our approach retains full resolution rendering, preserving best possible image quality, while achieving higher frame rates than present multi-view rendering methods. We describe our multi-view volume rendering algorithm and its implementation using programmable fragment shaders. Experimental results demonstrate our algorithm's improvement compared to prior multi-view volume rendering solutions.