Packet losses decrease the quality of an image or video for multimedia applications. Robust image coding is crucial to combat packet losses, for transmission of images over non feedback networks. New CS based image coding schemes are robust against packet losses and carries CS samples of nearly equal importance. CS based coding also ensures low costs and complexity for image sensing. Hence CS based image coding techniques have some distinct advantages over traditional Forward Error Correction (FEC) techniques and Multiple Description Coding (MDC) based methods . Forward error Correction techniques are generally employed along with some transform based coding , but provides a limited error resilience. MDC methods are considered to be one of the widely used mechanisms for packet losses. Compressive sensing based methods are an alternate to MDC and are able to provide robust image coding against packet losses with large number of descriptions. Recent work takes CS as a framework and Multiple Description Coding is done to get robust image coding against packet losses. The aim of the paper is to give a brief introduction to all the above techniques and survey four different CS based image coding techniques.

Packet reordering (RO) is an Internet event that degrades the performance of both TCP and UDP-based applications. In this paper, we present an end-to-end measurement study of packet reordering of UDP traffic. The goal of our measurement study is to characterize packet reordering in the current Internet as it is reflected by PlanetLab infrastructure. Overall, our analysis shows that current UDP traffic reordering is consistent to prior 1990's studies, despite increased Internet load and technology advancements. In addition, our study adds to the previous results by identifying additional reordering characteristics. More specifically, we show that packet reordering is asymmetric as well as temporal and site-dependent, packet size does influence the likelihood of reordering, that there exists a time-of-the-day dependency, and reordering primarily exists at two timescales (a few milliseconds or multiple tens of milliseconds.)

The high demand for large scale storage capacity calls for the availability of massive storage solutions with high performance interconnects. Although cluster file systems are rapidly improving and have the potential to allow extremely large numbers of commodity storage nodes to be pooled into a single large file-system, the number of ports on individual switches has not been increasing as quickly - the largest switches available today support fewer than 2,000 Gigabit Ethernet ports. Our goal, therefore, is to develop a new interconnect topology that can connect hundreds of thousands of nodes and achieve performance comparable to a single switch of equivalent size. At the same time, such a new topology should be readily buildable using inexpensive components. Our proposed architecture exploits the multiple Ethernet ports that are now standard on servers and combines host- based routing and forwarding with network-based switching to allow massively large storage clusters to be built. Simulation results have shown that our proposed design achieves 72% to 90% of the performance of a single switch capable of accommodating all storage nodes, but our approach scales to hundreds of thousands of nodes. Furthermore, we use common off-the-shelf layer-2 switches rather than more expensive models that support layer-3 routing. Finally, our approach is resilient to network faults because it maintains multiple paths between storage nodes.

In this paper we argue that a robust incentive mechanism is important in a real-world peer-to-peer streaming system to ensure that nodes contribute as much upload bandwidth as they can. We show that simple tit-for-tat mechanisms which work well in file-sharing systems like BitTorrent do not perform well given the additional delay and bandwidth constraints imposed by live streaming. We present pre- liminary experimental results for an incentive mechanism based on the Iterated Prisoner's Dilemma problem that al- lows all nodes to download with low packet loss when there is sufficient capacity in the system, but when the system is resource-starved, nodes that contribute upload bandwidth receive better service than those that do not. Moreover, our algorithm does not require nodes to rely on any information other than direct observations of its neighbors' behavior to- wards it.

In this paper, we present Chainsaw, a p2p overlay multicast system that completely eliminates trees. Peers are notified of new packets by their neighbors and must explicitly request a packet from a neighbor in order to receive it. This way, duplicate data can be eliminated and a peer can ensure it receives all packets. We show with simulations that Chainsaw has a short startup time, good resilience to catastrophic failure and essentially no packet loss. We support this argument with real-world experiments on Planetlab and compare Chainsaw to Bullet and Splitstream using MACEDON.

We explore Chord-like ring structures for Distributed Hash Tables (DHTs) and show that the Postage Stamp Problem (PSP) is equivalent to finding optimal structures for such ring topologies. We then describe a variant of the PSP that corresponds to ring-like DHTs that use greedy routing and develop an algorithm that smoothly trades off between the number of finger pointers and network diameter. We provide a dynamic programming solution to the number of nodes that can be supported as a function of the number of finger pointers and the network diameter and also note an interesting link to the Fibonacci sequence.

In this paper, we present the design of a credit-based trad- ing mechanism for peer-to-peer file sharing networks. We divide files into verifiable pieces; every peer interested in a file requests these pieces individually from the peers it is connected to. Our goal is to build a mechanism that supports fair large scale distribution in which downloads are fast, w ith low startup latency. We build a trading model in which peers use a pairwise currency to reconcile trading differences with each other and examine various trading strategies that peers can adopt. We show through analysis and simulation that peers who contribute to the network and take risks receive the most benefit in return. Our simulations demonstrate that peers who set high upload rates receive high download rates in return, but free-riders download very slowly compared to peers who upload. Finally, we propose a default trading strategy that is good for both the network as a whole and the peer employing it: deviating from that strategy yields litt le or no advantage for the peer.

We study the application of unequal loss protection (ULP) algorithms to motion-compensated video over lossy packet networks. In particular, we focus on streaming video applications over the Internet. The original ULP framework applies unequal amounts of forward error correction to embedded data to provide graceful degradation of quality in the presence of increasing packet loss. In this letter, we apply the ULP framework to baseline H.263, a video compression standard that targets low bit rates, by investigating reorderings of the bitstream to make it embedded. The reordering process allows a receiver to display quality video, even at the loss rates encountered in wireless transmissions and the current Internet.

We present two methods for protecting a region of interest (ROI) in a compressed medical image transmitted across a lossy packet network such as the Internet or a wireless channel. We begin with a high quality wavelet-based coder, the Set Partitioning in Hierarchical Trees (SPIHT) algorithm, which orders data progressively by coding the globally important information first. We then compress the ROI to a higher quality than the rest of the image by scaling the wavelet coefficients corresponding to the ROI. This approach moves ROI information earlier in the bit stream. Finally, we add more redundancy to the ROI than to the rest of the image by two techniques. With MD-SPIHT, we repeat wavelet coefficient trees corresponding to the ROI and code them to higher bit rates than the background trees. With ULP-FEC, we use forward error correction (FEC) in an unequal loss protection framework. We find that both methods increase the probability of receiving high quality ROI in the presence of packet loss.

We consider history independent data structures as proposed for study by Teague and Naor [3]. In a history independent data structure, nothing can be learned from the representation of the data structure except for what is available from the abstract data structure. We show that for the most part, strong history independent data structures have canonical representations. We also provide a natural less restrictive definition of strong history independence and characterize how it restricts allowable representations. We also give a general formula for creating dynamically resizing history independent data structures and give a related impossibility result.

We show that the problem of optimal bit allocation among a set of independent discrete quantizers given a budget constraint is equivalent to the multiple choice knapsack problem (MCKP). This result has three implications: first, it provides a trivial proof that the problem of optimal bit allocation is NP-hard and that its related decision problem is NP-complete; second, it unifies research into solving these problems that has to-date been done independently in the data compression community and the operations research community; third, many practical algorithms for approximating the optimal solution to MCKP can be used for bit allocation. We implement the GBFOS, Partition-Search, and Dudzinski-Walukiewicz algorithms and compare their running times for a variety of problem sizes.

Uniform Bit Loss Alexander E. Mohr Eve A. Riskin University of Washington Richard E. Ladner Abstract We explore the distribution of bits among wavelet image subbands that are transmitted across a channel with uniform bit loss, which models interleaved data sent across a packet erasure channel. We develop a closed-form expression relating the source rate-distortion curves to the rate-distortion curves that include the e ect of expected bit losses. We use that expression to determine optimal bit assignments for each subband. Our results show that the bit allocation is biased against long codeword indexes; as a result, the bit allocation is more uniform among subbands in the presence of loss than under conditions of no loss. To demonstrate our methods, we use a 3-level discrete wavelet transform and encode each subband with its own balanced tree-structured vector quantizer of block size 2 2. 1 Approach The wavelet transform is a linear orthonormal transform that concentrates signal energy in a small number of subbands, and any error introduced in the transform domain is quantitively identical in the signal domain. A tree-structured vector quantizer is progressive: a crude representation is transmitted rst, followed by better and better approximations in subsequent passes. Our approach to handling the loss of a bit in a codeword is to truncate it to the previously-received bits, thus treating those bits as an index to an internal node in the tree. 2 Adjusted Distortion We derive an equation for the expected distortion in the presence of bit loss, as seen by the receiver. Let r be the rate in bits per vector. Let Dr be the distortion at rate r. This work was supported by U.S. Army Research O ce grant DAAH04-96-1-0255, an NSF Young Investigator Award, and a Sloan Research Fellowship. Departments of Electrical Engineering and Computer Science and Engineering, University of Washington, Box 352500, Seattle, WA 98195-2500 B B D

We consider the problem of error control for receiver-driven layered multicast of audio and video over the Internet. The sender injects into the network multiple source layers and multiple channel coding (parity) layers, some of which are delayed relative to the source, Each receiver subscribes to the number of source layers and the number of parity layers that optimizes the receiver's quality for its available bandwidth and packet loss probability. We augment this layered FEC system with layered pseudo-ARQ. Although feedback is normally problematic in broadcast situations, ARQ can be simulated by having the receivers subscribe and unsubscribe to the delayed parity layers to receive missing information. This pseudo-ARQ scheme avoids an implosion of repeat requests at the sender and is scalable to an unlimited number of receivers, We show gains of 4-18 dB on channels with 20% loss over systems without error control and additional gains of 1-13 dB when FEC is augmented by pseudo-ARQ in a hybrid system, Optimal error control in the hybrid system is achieved by an optimal policy for a Markov decision process.

This paper describes an algorithm that achieves an approximately optimal assignment of forward error correction to progressive data within the unequal loss protection framework. It first finds the optimal assignment under convex hull and fractional bit allocation assumptions. It then relaxes those constraints to find an assignment that approximates the global optimum. The algorithm has a running time of O(hNlogN) where h is the number of points on the convex hull of the source's utility-cost curve and N is the number of packets transmitted.

We consider the problem of joint source/channel coding of real-time sources, such as audio and video, for the purpose of multicasting over the Internet. The sender injects into the network multiple source layers and multiple channel (parity) layers, some of which are delayed relative to the source. Each receiver subscribes to the number of source layers and the number of channel layers that optimizes the source-channel rate allocation for that receiver's available bandwidth and packet loss probability. We augment this layered FEC system with layered ARQ. Although feedback is normally problematic in broadcast situations, ARQ is simulated by having the receivers subscribe and unsubscribe to the delayed channel coding layers to receive missing information. This pseudo-ARQ scheme avoids an implosion of repeat requests at the sender, and is scalable to an unlimited number of receivers. We show gains of up to 18 dB on channels with 20% loss over systems without error control, and additional gains of up to 13 dB when FEC is augmented by pseudo-ARQ in a hybrid system. The hybrid system is controlled by an optimal policy for a Markov decision process.

We consider the problem of error control for receiver-driven layered multicast of audio and video over the Internet. The sender injects into the network multiple source layers and multiple channel coding (parity) layers, some of which are delayed relative to the source. Each receiver subscribes to the number of source layers and the number of parity layers that optimizes the receiver's quality for its available bandwidth and packet loss probability. We augment this layered FEC system with layered pseudo-ARQ. Although feedback is normally problematic in broadcast situations, ARQ can be simulated by having the receivers subscribe and unsubscribe to the delayed parity layers to receive missing information. This pseudo-ARQ scheme avoids an implosion of repeat requests at the sender and is scalable to an unlimited number of receivers. We show gains of 4-18 dB on channels with 20% loss over systems without …

We present the unequal loss protection (ULP) framework in which unequal amounts of forward error correction are applied to progressive data to provide graceful degradation of image quality as packet losses increase. We develop a simple algorithm that can find a good assignment within the ULP framework. We use the set partitioning in hierarchical trees coder in this work, but our algorithm can protect any progressive compression scheme. In addition, we promote the use of a PMF of expected channel conditions so that our system can work with almost any model or estimate of packet losses. We find that when optimizing for an exponential packet loss model with a mean loss rate of 20% and using a total rate of 0.2 bits per pixel on the Lenna image, good image quality can be obtained even when 40% of transmitted packets are lost.

and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the nal examining committee have been made. Date: In presenting this thesis in partial fullllment of the requirements for a Master's degree at the University o f W ashington, I agree that the Library shall make its copies freely available for inspection. I further agree that extensive copying of this thesis is allowable only for scholary purposes, consistant with fair use" as prescribed in the U.S. Copyright Law. Any other reproduction for any purpose or by any means shall not be allowed without my written permission. This thesis presents a framework in which images can betransmitted over channels with high packet loss rates with the addition of unequal amounts of forward error correction FEC. It develops an algorithm that can optimize the allocation of FEC to the image data so as to maximize the expected image quality. By using both the importance of the output of the image coder and an estimate of the probability of losing packets, it considers distortion-rate tradeoos in its assignments to provide graceful degradation when packets are lost. The algorithm is also modular in that it can use any compression scheme that produces a progressive bitstream. 2.1 In Leicher's application of PET to MPEG 111, he applied 60 priority to message fragment M 1 the I frames, 80 priority to M 2 the P frames, and 95 priority t o M 3 the B frames. Each message and its associated FEC use the same range of bytes in every packet.. .. .. 9 3.1 Each of the rows is a stream and each of the columns is a packet. A stream contains one byte from each packet.