Fast 2-Dimensional 8x8 Integer Transform Algorithm Design for H.264/AVC Fidelity Range Extensions

SUMMARY In this letter, efficient two-dimensional (2-D) fast algorithms for realizations of 8 × 8 forward and inverse integer transforms in H.264/AVC fidelity range extensions (FRExt) are proposed. Based on matrix factorizations with Kronecker product and direct sum operations, efficient fast 2-D 8 × 8 forward and inverse integer transforms can be derived from the one-dimensional (1-D) fast 8 × 8 forward and inverse integer transforms through matrix operations. The proposed fast 2-D 8 × 8 forward and inverse integer transform designs don't require transpose memory in hardware realizations. The fast 2-D 8 × 8 integer transforms require fewer latency delays and provide a larger throughput rate than the row-column based method. With regular modularity, the proposed fast algorithms are suitable for VLSI implementations to achieve H.264/AVC FRExt high-profile signal processing.