Uyarlanır ölçeklenebilir video kodlama

This thesis is composed of three main parts which include three contributions in slightly differentfields, all lying on the same framework: Adaptive Scalable Video Coding. First part is aboutintegration of motion compensated temporal filtering (MCTF), the basis for temporal scalabilityin scalable video coding methods, to the latest non-scalable video compression standard, i.e.,H.264/AVC. We propose a GOP structure to implement block-based adaptive MCTF within theH.264/AVC syntax using stored B-pictures, similar to the motion-compensated 5/3 waveletfiltering. We provide experimental results to compare the results of our proposed codec withthose of other scalable wavelet video coders which use MCTF. The proposed scheme is alsointegrated into H.264/AVC reference software as `Hierarchical B pictures? or `TemporalPyramid? and it is currently under investigation of MPEG Core Experiments for the upcomingScalable Video Coding standard (SVC).Secondly, we worked on content adaptive scalability type selection problem. State of the artscalable video coders provide different options, such as temporal, spatial and SNR scalability,where bitrate reduction using each scalability type results in different kinds and/or levels of visualdistortion depending on the content and the bitrate. In most cases, a single scalability type doesnot fit the whole video well, and scaling option selection can be optimized for each temporalsegment depending on the content of that segment and the target bitrate. This dependencybetween selection of scalability type, video content, and bitrate is not well investigated in theliterature. In this work, assuming that the video is temporally segmented by some content analysisscheme, we propose a method to choose the best scaling type for each temporal segment thatresults in minimum visual distortion among temporal, spatial and SNR scalability for fullyembedded scalable video coders. We employ an objective distortion measure that consists of alinear combination of four component measures, which are a flatness measure, a blockinessmeasure, a blurriness measure, and a temporal jerkiness measure, to quantify artifacts caused bybitrate reduction by spatial size reduction, frame rate reduction, and quantization parameterscaling. Two subjective tests have been performed to validate the proposed procedure for shot-based selection of optimal scalability type on soccer videos. Soccer videos whose bitrate arereduced from 600 kbps to 100-300 kbps by the proposed content-adaptive selection of scalabilitytype have been deemed visually superior to those whose bitrates are reduced by a singlescalability option for the entire test sequence.Finally, we worked on adaptive peer-to-peer (P2P) streaming using scalable multiple descriptioncoding. Efficient P2P video streaming is a challenging task due to time-varying nature of both thenumber of available peers and network/channel conditions. To this effect, we propose i) a newflexible scalable multiple description coding (MDC) method, where the number of descriptions,and the rate and redundancy level of each description can be adapted on the fly (by post-processing of a fully-embedded scalable coded bitstream), and ii) a new adaptive TCP FriendlyRate Controlled (TFRC) P2P streaming system based on this new MDC scheme. Theoptimization of the design parameters of the proposed MDC scheme according to networkconditions is discussed within the context of the proposed adaptive P2P streaming framework,where the number and quality of available streaming peers/paths are a priori unknown and vary intime. Experimental results, by means of NS-2 network simulation of a P2P video streamingsystem, show that adaptation of the number and rate of descriptions/layers and the redundancylevel of each description according to network conditions yields significantly superiorperformance when compared to other scalable MDC schemes using a fixed number ofdescriptions/layers with fixed rate and redundancy level.