Statistical analysis and parametrization of HEVC encoded videos

摘要

HEVC is planned to provide better coding efficiency. However, this efficiency comes with an important computational complexity. Many works are undergoing to decrease the complexity of the encoder. Nevertheless, the analysis of the encoder' decision is necessary to have an idea about partitioning mode of the CUs in terms of size and type. This paper presents statistical analyses of coding units (CU) chosen by the encoder. These statistics are followed by a texture analysis. A texture parameter was established for the different videos encoded using HEVC. The analysis of the texture is made using the Sobel filter. The results show that the percentage of choosing the coding units depends directly on the video characteristics. In fact, for textured videos, having a lower texture parameter, there is a need to process with smallest coding units. This work is as milestone for proposing algorithms based on video characteristics to perform fast partitioning.