Contents of this page are obsolete. This page is preserved and stored at this URL just from historical viewpoint. Original URL was http://www.mm.media.kyoto-u.ac.jp/members/kameda/...
Please visit www.kameda-lab.org for recent information. (2002/12/06, [email protected])

Publication, Research, Docs and Info, KAMEDA


next up previous
Next: Experiment Up: Multi-user live video Previous: Video imaging for

Mediation and user satisfaction

Request of an user j is satisfied if a camera-work in is selected and assigned to a certain camera at time t. Suppose there are u users and c cameras. As a camera can realize only one camera-work at a time, at most c camera-works can be realized at time t. As a consequence, mediation for all the user requests is necessary to select at most c camera-works among where . Let us denote be the summed number of the camera-works in for all j. If it includes the same camera-works, they are counted as one camera-work. The mediation is described by the mediation matrix of rows and u columns where each component is either 1 or 0. A column represents and a row corresponds one camera-work i. means that the user j realizes his/her request of the camera-work i. Hence, the next inequality ought to be true for all j.

 

Note that at each column j, is always 0 if camera-work i is not included by . We introduce that indicates whether the corresponding camera work of the ith row is selected or not. indicates the number of the users who support the camera-work i.

 

 

The constraint of the number of the cameras is formulated by

 

The mediation process is finding the mediation matrix M that maximizes under the constraints of Inequality (2) and (5).

where is defined as:

 

The mediation matrix that satisfies is always found in the case , but there might be a dynamic situation where u > c and so .

We use greedy algorithm to solve this problem. Although it does not give us the best solution that maximizes , it reaches locally optimal solution with less amount of calculation, which is essential in real-time processing.

After the camera-works are selected, the next problem is to assign the camera suitable to each selected camera-work under the constraint of the camera location, direction and range. The criteria to solve this problem is to satisfy the requests of the users as much as possible. Our strategy is as follows.

  1. Order the selected camera-works by .
  2. Choose the best camera for each camera-work in the order.

In this way, the video image the user j watches is taken by assigning the camera the selected camera-work in the mediation matrix if his/her request is realized.

On the other hand, there might be a case where some users cannot find their camera-works in the selected camera-works.

In this case, the user watches an arbitrary chosen video image.



next up previous
Next: Experiment Up: Multi-user live video Previous: Video imaging for



Yoshinari Kameda
Fri Oct 1 16:26:35 JST 1999