Locations of visitors to this page

2.7 million Key NSFC project granted


The Multimedia Lab was granted a Key NSFC for International Cooperation, entitled "Object and MacroCube-based Coding and Understanding 3D Video". The project was co-applied by multimedia lab of XJTLU in collaboration with researchers from the Beijing Jiaotong University and Taiyuan University of Science and Technology and the University of Torino (Italy).

The project won a research fund of 2.7 million Yuan, half of which will be granted to support Dr. Tillo and his PhD research team. This project belongs to the category of Key NSFC project, which is more difficult to apply than the General Programme and it is the first time that XJTLU is awarded this kind of project. The success application of this project will help to further strengthen the research at the Multimedia Lab.

Due to the recent advances in video capturing and displaying technologies, the 3-D video communication and entertainment services are expected to become reality, and consequently the multimedia research community is focusing more and more on redefining the communication and entertainment services to suit the need of this emerging field. However, most of the research work related to compression, processing, and exploitation of the 3-D data is built over the conventional 2-D paradigms. In this project, we aim to change this paradigm by investigating, proposing, designing, and testing different set of mechanisms which work directly in the 3-D space. This is because we believe that the simple usage and adaptation of 2-D video concepts to the emerging 3-D data will end up scarifying the potential of the third dimensionality.

The first think we want to do in this project is to have a new way of representing the 3-D data. Given that the data we want to deal with is 3-D data we think that representing the data directly into a 3-D format will be more convenient, and promising. In this case, the data acquisition stage will be separated from the data processing stage, so we will have a universal way to describe the 3-D data which is independent from the way the original 3-D video is generated or acquired. Because the 3-D content, if well represented and exploited, carrys a lot of useful information, then we will investigate how to exploit this information so as to have better understanding of the data and the objects in the scene. So in order to obtain high coding efficiency, we will investigate high-level descriptors of the moving objects and background, we aim to propose algorithms to identify the moving objects in the 3-D space, and to describe their luminance and movement as a rotated/scaled/translated version of their previous version. Whereas, for the subtle moving objects we will extend the concept of 2-D motion estimation to the 3-D space, so we will investigate the 3-D motion estimation of MacroCube (cuboid), instead of the conventional MacroBlock, moreover, we will investigate 3-D transforms to de-correlate the motion compensated 3-D frame. Moreover, in this project we will devise some algorithms to allow having Free Viewpoint Video from the compressed stream to enable seamless interactivity, where a user could simply extract the views he/she wants from the received stream with a limited complexity at the client side. Moreover, object categorization and classification will be another task that we will address. This will increase the usability of the data while reducing the complexity of the processing stages. Furthermore, we will investigate techniques to allow extracting object of interest. All the issues addressed in this proposal are indeed important to allow the adaptation of the new 3-D services into the current distribution/storage/broadcasting platforms. For these reasons, we believe that the outcomes of this project are potentially relevant to several standardization bodies.

ScrewTurn Wiki version Some of the icons created by FamFamFam. 苏ICP备11062770号
Mailbox Edit Login