Researches On Video Adaptation Based On Multidimensional Description

Video adaptation has been considered as an effective technique to tackle challenging problems in pervasive media applications over heterogeneous networks and diversified end-users.It transforms an inputed video to an outputted video in a new format or an augmented multimedia form to meet diverse resource constraints and user preferences.It is a promising technique in theory and practice to solve the problems of UMA(Universal Multimedia Access).Our research focuses on the universal video adaptation framework based on multidimensional description,and the key techniques including multidimensional description design,adaptation operation realization,and utility based operation selection.The innovations of this dissertation are as follows:1) Propose a video adaptation scheme based on multidimensional descriptionTo give an example of the universal framework based on multidimensional description,this dissertation proposes a practical scheme to video adaptation, considering the requirements of ubiquitous video application environment.Motion information,structural information and texture information are depicted in multidimensional description,so that the solution can tackle a number of different adaptation applications in not only signal level,but also structural level and joint level.2) Propose a new solution to video format adaptation based on motion descriptionDue to the co-existence of various video coding standards in multimedia applications,it is of great significance to implement video format adaptation among these standards.The dissertation presents a novel solution to video format adaptation according to the universal framework based on multidimensional description. Detailed and complete motion description generated by H.264/AVC pre-encoding is utilized to fast encoding or transcoding the bit-stream into different video format and bit-rate.3) Propose a structure-aware hierarchical motion descriptionA structure-aware hierarchical motion description is proposed as an important element of multidimensional description.It integrates motion and structural information of video signal,which enables multiple adaptation operations not only in signal level(e.g.,bit-rate conversion and temporal resolution reduction) but also in structure level(e.g.,video summarization and extraction of interesting segments) and joint level.4) Propose a video texture compression algorithm with high qualityA new texture compression method for video sequence(YUV 4:2:0) with high quality is proposed to generate texture description for multidimensional description. The idea is borrowed from conventional texture compression method for RGB image termed DXTC(DirectX texture Compression),which is improved to enhance the compression quality.Meanwhile,the compressed texture can be used together with the structure-aware hierarchical motion description mentioned above for adaptation applications without original video sequence.The author conducted plenty experiments to verify that the proposed multidimensional description can support the video adaptation operations in real-time applications,while maintaining coding efficiency.The results demonstrate the feasibility and effectiveness of universal video adaptation framework based on multidimensional description which worth further researches.