Abstract
Recently, there has been an increasing demand to develop efficient compression algorithms for high definition videos. Both motion estimation and motion compensation play important roles in the video coding framework. The application of motion based pre-processing and advanced motion compensation technology could further improve video compression performance. This thesis focuses on the performance improvement and transcoding optimization for the latest video coding standards including VVC and ECM. It mainly consists of three parts: 1) Enhanced motion compensated temporal filtering; 2) Efficient interpolation filters for chroma motion compensation; 3) Low-complexity transcoding from HEVC to VVC based on HEVC motion information and mode information; The first part aims to improve the RD (Rate-Distortion) performance of video coding by introducing multi-hypothesis references and adaptive weights based on motion variance and texture complexity in MCTF. The second part is to improve the performance of chroma motion compensation by using 6-tap chroma interpolation filter. The last part is to accelerate HEVC-to-VVC transcoding by using HEVC motion information and mode information, which is beneficial to the industrial transcoding application of VVC.In the first part, an enhanced motion compensated temporal filtering based on multi-hypothesis reference is proposed, with the consideration of motion vector variance, and texture complexity. We take initial steps towards the incorporation of motion vector variance and texture complexity in the filtering weights design. Motion compensation blocks based on multi-hypothesis reference, can be efficiently aggregated in an effort to obtain the final inference. Experimental results show that promising coding gain can be achieved with little additional encoding complexity increase.
In the second part, a 6-tap DCT based chroma interpolation filters for chroma motion compensation is studied by comprehensively analyzing different types of interpolation filters. The proposed interpolation filters are designed considering multiple factors, comprising the characteristics of the chroma component, the frequency response of interpolation filter and the expenses of computational complexity. The experimental results show that the proposed chroma interpolation filter significantly improve the RD performance of video coding with small decoding time increasing. Owing to the good trade-off between coding performance and complexity, the proposed method has been adopted in ECM software.
In the third part, a low-complexity HEVC to VVC transcoding scheme is investigated. The partition and motion information in HEVC bitstream implies the intrinsic spatial and temporal characteristics of the video sequence. Therefore we propose a new solution for HEVC-to-VVC transcoding by combining the mode information from HEVC and the intermediate coding information from VVC. Additionally, motion information from HEVC bitstream is used to prune the decision process of CU partition, and the mode information from HEVC bitstream is used to skip the mode decision of VVC. Experimental results show the proposed scheme significantly reduce the transcoding complexity with small RD performance loss.
Therefore, this thesis studies enhanced video coding technologies, which contributes to the application of new generation coding standards and the compression, storage and transmission of high-quality video data. Specifically, the characteristics of pre-analyzed motion information and the implementation of motion compensation technology are systematically studied to improve the performance of video compression. This Thesis objectively evaluates the performance of the proposed method, which will promote the industrial application of video coding and reduce the cost of video data applications.
| Date of Award | 23 Aug 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Shiqi WANG (Supervisor) |