Abstract for the Project Content Based Image Retrieval By Active Learning SVM

Started by Kalyan, Aug 22, 2008, 08:54 PM

Previous topic - Next topic

Kalyan

Abstract for the Project Content Based Image Retrieval By Active Learning SVM

The project is focused on minimizing the time taken for the buffering operation of multimedia files. The emphasis is mainly on balancing the load of the computing servers that performs the streaming and transcoding of the multimedia file.

The media file is clustered into streams of equal size and it is stored in frames for easy retrieval. The load of each computing server is calculated with the help of the token generation. The number of tokens generated will be proportional to the remaining load of the computing server.

Using the Least Load First Algorithm, the least load proxy is identified. The frames are dispatched to the least load computing server after performing the necessary scheduling operation. The proxies within the same region will share the information respectively.

The cache of each proxy will hold the media files that are streamed in prior. If another lient happens to buffer the same media file through any other proxy within that region then that proxy will divert the call to the proxy that holds the required media file. But this process is a rare real time situation. We further extend the project to handle real time situation where different clients requests different media files.

With the help of the tokens generated we will be able to dispatch the frames to the appropriate proxy for streaming. The performance is improved since multiple proxies are used for the streaming of a media file rather than a single proxy in the case of the existing system. Hence the load of all the available computing servers are utilized efficiently, thus minimizing the total buffering time.