Multi-view Rendering using GPU for 3-D Displays

Abstract

Creating computer graphics based content for stereoscopic and auto-stereoscopic displays requires to render a scene several times from slightly different viewpoints. In that case, maintaining real-time rendering can be a difficult goal if the geometry reaches thousands of triangles. However, similarities exist among the vertices belonging to the different views like the texture, some transformations or parts of the lightning. In this paper, we present a single pass algorithm using the GPU that speeds-up the rendering of stereoscopic and multi-view images. The geometry is duplicated and transformed for the new viewpoints using a shader program, which avoid redundant operations on vertices.