Fast 8×8 DCT pruning algorithm

Most of hybrid motion compensated video coding standards uses a well known discrete cosine transform (DCT) at the encoder to remove redundancy from video random processes. Due to the huge number of computations required to compute the DCT, reduction of its complexity is essential to speed up video processing. To optimize the computation of the DCT, we take into account the fact that the output of the DCT has often small values. The goal is not to compute DCT coefficients that will be quantized towards zero. This method of computing only a subset of DCT coefficients is commonly referred as pruning. In this paper we propose a method to perform the pruning of the two-dimensional (2-D) DCT based on the algorithm proposed in E. Feig and S. Winograd (1992) for any number of low-frequency components.