Compression of the Lumigraph/Lightfield

Figure 1. 2D image array of Lumigraph/Lightfield

A Lumigraph/Lightfield captures the complete appearance of a 3D object/scene in the form of 2D image array, and is able to quickly render an arbitrary view independent of the scene/object complexity. Since the data amount of the Lumigraph is huge, the efficient storage and access of the Lumigraph/Lightfield is essential. In this work, we propose a multiple reference frame (MRF) structure to compress the Lumigraph/Lightfield. By predicting the Lumigraph/Lightfield images from the neighbor views and compressing the prediction residue, highly efficient compression is achieved. More important, the MRF is able to achieve the functionality of just-in-time (JIT) rendering, in which only the portion of the compressed bitstream needed to render the current view is accessed, decoded and rendered. Compared with first expanding the compressed Lumigraph and then rendering, JIT rendering eliminates the need of a huge buffer at the decoding end, and greatly reduces the cost of the rendering engine. The decoder cache is used extensively in the MRF render to speed up rendering. The trade off between the computational speed and memory requirement of the MRF render is discussed.

For more details of the work, please refer to the followings:

©Copyright by Jin Li, June 22, 2001