Youtube*

4 September 2009 (Friday) ‘s seminar was interactive and useful. We learnt a lot from Dr Ooi Wei Tsang about how YouTube manages the videos uploaded by users worldwide. In that seminar, he discussed about video compression, video distribution, video thumbnails, video fingerprinting, video spammers and video recommendation. Below are a few highlights on what was discussed…

What is video compression?
As many know, videos take up a lot of space since 1 pixel= 1 byte for red + 1 byte for blue + 1 byte for green and thus, it has to be compressed before they are being put up on web. In the seminar, our speaker mentioned that one way of video compression is lossy compression. This type of compression will however cause the compressed video file to have less data and this may affect the quality of the video. Another way would be to make use of temporal redundancy where we can reduce data by looking for pixels in two video frames that have the same values in the same location.

How videos are distributed?
Videos uploaded to YouTube are stored in a system of computers, placed at various places in the world so that user can download videos from the server that is near their location instead of from the central server so as to minimize a bottleneck near the central server. This system of computers is called the content delivery network (CDN) as mentioned by our speaker. He also mentioned that limelight provides the content delivery network service (www.limelightnetworks.com).

What is video fingerprinting?
Video fingerprinting is a technique that helps to prevent users to upload videos that infringe the copyright laws. As the word ‘fingerprinting’ suggests, this technique will help to check whether the video uploaded by the user has unique images by comparing with a library of materials that are marked as copyrighted. Video fingerprinting analysis may be based on key frame analysis and the colour and motion changes during a video sequence.

Video thumbnails and Video recommendation
The thumbnail of a video is what users see when searching or browsing YouTube videos. These thumbnails will provide some screen shots of the content of the video so as to allow the users to decide which video they want. Video recommendation is done through filtering out those videos that are most commonly watched by people. Recommendations can be done by asking the user to rank the videos that they have watched. YouTube also has the active sharing feature which allows their watchers to signal to other viewers that they are watching the same video at a particular time.

In conclusion, the seminar on how YouTube works was useful as it provides me with lots of knowledge that I never knew and it also allows me to understand the ‘huge’ amount of work behind the scene of YouTube (from uploading of videos to distribution and preventing the infringement of copyright laws).

By:Geck Keat and Byron