A fast three dimensional discrete cosine transform

Abstract

Summary form only given. A three dimensional fast discrete cosine transform (3-D FCT) algorithm is proposed for 3-D data points. Unlike other methods for 3-D DCT, the proposed algorithm treats 3-D data points directly as volume data. The algorithm involves a 3-D decomposition and rearrangement process where a data volume is recursively halved for each dimension until unit data cubes are formed. The data points are further rearranged to avoid redundant computations in the transformation. The 3-D algorithm has been shown to be computationally efficient, and can thus be used in applications requiring a real-time symmetric codec, such as software based video conferencing systems.