FAST MOTION ESTIMATION ALGORITHM FOR HEVC VIDEO CODEC

Subject of Research. The paper considers motion estimation as the most difficult and resource-consuming process in the high efficiency video coding standard. A fast algorithm is proposed that skips the fractional-pixel motion estimation. Method. The fractional-pixel motion estimation, including the half-pixel and quarter-pixel motion estimation, is used as a refinement process for integer-pixel motion estimation to provide better coding efficiency. The proposed algorithm involves skipping the fractional motion estimation for prediction blocks if the block with 2N × 2N dimension, which contains more information than the other block sizes, is encoded with integer motion estimation. The decision can also be made in reference frames, so that motion estimation at half- and/or quarter-pixel levels can be skipped to reduce encoding time. Main Results. The proposed algorithm is implemented in HM-16.20 reference software. Several test sequences recommended by the International Telecommunication Union (ITU) are used to evaluate its effectiveness. Experimental results show that the proposed fast algorithm reduces the encoding time by an average of 64 % with a performance loss of less than 1 %. Practical Relevance. The proposed fast algorithm provides a reduction in coding time with a slight loss in performance. It can replace an existing unit in the H.265 standard speed up the video compression process.