Unified and fast 2-dimensional 4/spl times/4 transform design for H.264/AVC texture coding

In this paper, the unified two-dimensional (2-D) fast algorithms for realization of 4/spl times/4 integer and Hadamard transforms in H.264/AVC are proposed. Based on matrix operations with Kronecker product, the efficient fast 2-D 4/spl times/4 transforms can be derived from the proposed one-dimensional (1-D) fast 4/spl times/4 transforms through matrix decompositions. The proposed fast 2-D 4/spl times/4 transform design doesn't require transpose memory for unified architecture. The fast 2-D 4/spl times/4 transform design requires fewer latency delays than the state-of-the-art method. With Xilinx FPGA verifications, the proposed fast and unified design performs better hardware efficiency than the well-known 1-D based row-column architecture.