Abstract :
Wireless networks are very popular today; however, the bandwidth over the air interface is limited to support many real time applications like video on demand. While some of these applications require high quality picture frames with high resolution, large image size, etc., many others (like remote surveillance) need simple and continuous, yet intelligible image frames that are not jittery. A video server that has an estimate of the available bandwidth can stream high quality data if the bandwidth is available, or stream data that needs lower bandwidth (perhaps with a lower size, lower frame rate, etc.) if the available bandwidth is limited. Unfortunately, that poses two issues - first, the video server should be aware of the available bandwidth. Second, it should be able to modify the content it streams, "dynamically". In this paper, we do a detailed analysis of some popular "controlled lossy compression" methods that are used to reduce the overall "entropy" of the source data (video stream), leading to a smaller data size and reducing the overall bandwidth requirements. We discuss these methods in the context of image processing techniques and show how they can modify the data of stored video content, as needed, to present an intelligible video stream that is smaller is size. We present an end to end architecture that (i) uses "RTCP" to get an estimate of the available channel bandwidth and (ii) uses a video server that can dynamically change the parameters like image size and frame rate to produce video streams that are best suited for the available bandwidth. We present theoretical compression rates for each of these compression methods and present empirical results to confirm our claims.
