Authors: John Dingliana, Salaheddin Alakkari
Abstract: We present an image-based volume visualization approach based on Principal Component Analysis (PCA). Using PCA, a learned model is trained using pre-rendered images from spherically distributed viewing angles. The views are encoded into a compressed and more compact representation and then novel views can be synthesized by interpolating scores in the eigenspace. The main advantage of the PCA model is the low computational complexity in the encoding and decoding phases. Furthermore, the image encoding and reconstruction is independent of the rendering complexity. This is particularly important in the case of computationally demanding rendering techniques such as global illumination. Our technique has potential application in client-server volume visualization or where results of a computationally-complex 3D imaging process need to be interactively visualized on a display device of limited specification.