Volume Rendering Using Principal Component Analysis

Abstract

We investigate the use of Principal Component Analysis (PCA) for image-based volume rendering. We compute an eigenspace using training images, pre-rendered using a standard raycaster, from a spherically distributed range of camera positions. Our system is then able to synthesize novel views of the data set with minimal computation at run time. Results indicate that PCA is able to sufficiently learn the full volumetric model through a finite number of training images and generalize the computed eigenspace to produce high quality novel view images.