Interactive transmission and visually lossless strategies for JPEG2000 imagery

Abstract

Every day, videos and images are transmitted over the Internet. Image compression allows to reduce the total amount of data transmitted and accelerates the delivery of such data. In video-on-demand scenarios, the video has to be transmitted as fast as possible employing the available channel capacity. In such scenarios, image compression is mandatory for faster transmission. Commonly, videos are coded allowing quality loss in every frame, which is referred to as lossy compression. Lossy coding schemes are the most used for Internet transmission due its high compression ratios. Another key feature in video-on-demand scenarios is the channel capacity. Depending on the capacity a rate allocation method decides the amount of data that is transmitted for every frame. Most rate allocation methods aim at achieving the best quality for a given channel capacity. In practice, the channel bandwidth may suffer variations on its capacity due traffic congestion or problems in its infrastructure. This variations may cause buffer under-/over-flows in the client that causes pauses while playing a video. The first contribution of this thesis is a JPEG2000 rate allocation method for time-varying channels. Its main advantage is that allows fast processing achieving transmission quality close to the optimal. Although lossy compression is the most used to transmit images and videos in Internet, when image quality loss is not allowed, lossless compression schemes must be used. Lossless compression may not be suitable in scenarios due its lower compression ratios. To overcome this drawback, visually lossless coding regimes can be used. Visually lossless compression is a technique based in the human visual system to encode only the visually relevant data of an image. It allows higher compression ratios than lossless compression achieving losses that are not perceptible to the human eye. The second contribution of this thesis is a visually lossless coding scheme aimed at JPEG2000 imagery that is already coded. The proposed method permits the decoding and/or transmission of images in a visually lossless regime.

Date of Award	30 May 2014
Original language	English
Supervisor	Francesc Auli Llinas (Director)