Markov Chains Fusion for Video Scene Generation

In this paper we address the general issue of merging Markov chains used to model two instances of a given process with some properties in common. In particular, in this work we apply this scenario to a multimedia application that generates new video scenes mixing the original segments of a given movie. To perform the latter process, it is first necessary to describe the structure of the scenes in some way, which in our case is done through Markov chains. The video scenes are then recombined by fusing their corresponding models using the general method described here. We analyze and validate the proposed methodology only for this specific application, however the solution presented here could be used in a very diverse array of applications where Markov chains are routinely used, ranging from queuing modeling to financial decision processes.