Efficient architectures for HEVC luma interpolation filter

Fractional-pel interpolation for motion estimation/ compensation is one of the most computational consuming areas in High Efficiency Video Coding (HEVC). This work presents an efficient design and implementation for luma interpolation filter in terms of hardware complexity and throughput. A new scaling factor for luma interpolation filter is adopted. By applying this modification, remarkable improvement is accomplished on hardware complexity. We propose two different architectures with fewer adders. In addition, optimization is applied on adders´ bitwidth, significant improvement in area reduction is achieved; up to 40% compared to the best architecture in the literature.