Comparison of different parallel implementations for deblocking filter of HEVC

Unlike deblocking filter of H.264/AVC, deblocking filter of HEVC is computationally less complex and offers more parallelization possibilities. In this paper we present comparison of three different parallelization implementations of deblocking filter which operate on picture by picture basis. The first method filters the vertical edges and horizontal edges in separate passes. The other two methods combine the vertical and horizontal edge filtering in a single pass. On a 6 core machine with 6 threads running concurrently, experimental results showed an average accelerating factor of 4.5, 5 and 5 for each of the implementation methods on 1080p, 1600p and 2160p sequences respectively.