An Efficient Merge Mode with Motion Vector Difference Approach for Versatile Video Coding.

To address the problem in H.266/VVC that the newly proposed merge mode with motion vector difference (MMVD) cannot fully reflect the directional changes in inter prediction due to the increase of object flexibility, This paper proposes a directional improvement technique based on inter prediction of MMVD mode by expanding the number of directions included in the original technique from four to six. In order to reduce the computational complexity of the internal prediction of MMVD process and obtain better coding efficiency, the original step size of 32 precision was reduced to step size of 2 precision. The experimental results show that compared with the H.266/VVC standard algorithm, the algorithm in this paper can save 0.34% of BD-rate for luminance component Y and 0.30% of BD-rate for chrominance component U of the test sequence under the LDB configuration, while the coding and decoding time remains stable even decreasing under the step size.