Colloquium Computer Science, Ananda Shankar Chowdhury, Jadavpur University (India)

Title: Multiple Video Processing using Graph-theoretic Frameworks

Abstract:

With recent advancements in video capture resulting from sensor technologies and availability of storage and distribution facilities, extent of video content access has increased exponentially. It has been also observed that instead of a single video, multiple videos can be more meaningfully analyzed for the purpose of video surveillance. To utilize such huge amount of data, proficient video management systems need to be developed. In this lecture, I will show how important information can be extracted from multiple videos using graph-theoretic frameworks. In particular; I will address two different problems, namely, multi-view video summarization and causal multi-spectral video fusion, both of which can play significant roles in video surveillance. Graph-theoretic frameworks have become popular for image and video processing due to factors like proper representation of high-dimensional data, proven polynomial time-complexity, scalability and ability to handle clusters of different shapes. Multi-view video summarization problem, where the goal is to obtain a meaningful summary from multiple videos, will be solved using bipartite graph matching and optimal path forest clustering. The problem of causal multispectral video fusion deals with fusion of videos from both visible and infrared spectra where future frames may not be available at the time of processing the current frame. I will discuss a superpixel based pipeline with graph similarity driven saliency models to solve this problem.

Colloquium coordinators are Prof.dr. M. Aiello (e-mail : M.Aiello rug.nl ) and
Prof.dr. M. Biehl (e-mail: M.Biehl rug.nl )

http://www.rug.nl/research/jbi/news/colloquia/computerscience