科研成果详情

In this paper, we propose two different online approaches for synthesizing video textures. The first approach is through incremental Isomap and Autoregressive (AR) process. It can generate good quality results, nonetheless, it does not extend well for videos that are more sparse, such as cartoons. Our second online video texture generation approach exploits incremental spatio-temporal Isomap (IST-Isomap) and AR process. This second approach can provide more efficient and better quality results for sparse videos than the first approach. Here, the IST-Isomap, which we proposed, is an extension of spatio-temporal Isomap (ST-Isomap) and incremental Isomap. It contains spatio-temporal coherence in the data set and can also be applied in an incremental mode. Compared with other video texture generation approaches, both of our online approaches are able to synthesize new video textures incrementally which in turn offer advantages (e.g. faster and more efficient) in applications where data are sequentially obtained. © 2009 Springer-Verlag.