Highly parallel implementation of sub-pixel interpolation for AVS HDTV decoder

In this paper, we propose an effective VLS1 architecture of sub-pixel interpolation for motion compensation in the AVS HDTV decoder. To utilize the similar arithmetical operations of 15 luma sub-pixel positions, three types of interpolation filters are proposed. A simplified multiplier is presented due to the limited range of input in the chroma interpolation process. To improve the processing throughput, a parallel and pipelined computing architecture is adopted. The simulation results show that the proposed hardware implementation can satisfy the real-time constraint for the AVS HDTV （1 920× 1 088） 30 fps decoder by operating at 108 MHz with 38.18k logic gates. Meanwhile, it costs only 216 cycles to accomplish one macroblock, which means the B frame sub-pixel interpolation can be realized by using only one set of the proposed architecture under real-time constraints.

In this paper, we propose an effective VLSI architecture of sub-pixel interpolation for motion compensation in the AVS HDTV decoder. To utilize the similar arithmetical operations of 15 luma sub-pixel positions, three types of interpolation filters are proposed. A simplified multiplier is presented due to the limited range of input in the chroma interpolation process. To improve the processing throughput, a parallel and pipelined computing architecture is adopted. The simulation results show that the proposed hardware implementation can satisfy the real-time constraint for the AVS HDTV (1920×1088) 30 fps decoder by operating at 108 MHz with 38.18k logic gates. Meanwhile, it costs only 216 cycles to accomplish one macroblock, which means the B frame sub-pixel interpolation can be realized by using only one set of the proposed architecture under real-time constraints.