The Wavelet Stream: Interactive Multi Resolution Light Field Rendering

Abstract

One of the most general image based object representations is the LightField. Unfortunately, a large amount of data is required to reconstructhigh quality views from a Light Field. In this paper, we present thewavelet stream which employs non-standard four-dimensional waveletdecomposition for Light Field compression. It allows for progressivetransmission, storage, and rendering of compressed Light Field data. Ourresults show that 0.8% of the original coefficients or 0.3 bits perpixel, respectively are sufficient to obtain visually pleasing newviews. Additionally, the wavelet stream allows for an adaptivemulti-resolution representation of the Light Field data. Furthermore, asilhouette-encoding scheme helps to reduce the number of coefficientsrequired. Our data structure allows to store arbitrary vector-valueddata like RGB- or YUV-data. The Light Field data stored in the waveletstream can be decompressed in real time for interactive rendering. Forthis, the reconstruction algorithm uses supplementary caching schemes.