Jiang W. and Ortega A., Multiple Description Coding via Polyphase Transform and Selective Quantization, in Proceedings of Visual Communications and Image Processing, VCIP'99, (San Jose, USA), January 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....due to losses by providing class based degree of protection. Similar to the resynchronization techniques, the best results are however obtained only with a judicious packet prioritization process [18] 19] In this category, layered coding [20] 21] and the Multiple Description Coding schemes [22] can be mentioned as the most promising algorithms. Optimal error resilient schemes should, however, not only combine techniques of the above three categories, but also exploit the local relevance of video data. Given bit budget constraints, such a combination is indeed the only way to provide ....

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proc. Visual Communications and Image Processing, VCIP'99, San Jose, CA, Jan. 1999.

....set of descriptions with some redundancy. In this case, the redundancy is deterministic and is caused by using a larger expansion set than is necessary to form a basis. Data partitioning is another common technique used to add error resiliency to a source coder, and examples include the work in [38, 71, 90, 91]. High performance compression algorithms, such as the SPIHT image coder [10] often require the source coder bits to be decoded in a sequential and uninterrupted fashion for correct interpretation. Carefully partitioning the data into separate streams prior to coding can improve the error ....

....in the higher frequency subbands. For these situations, data partitioning can be applied so that information about a specific high frequency coefficient is only sent in a subset of the packets. A similar type of data partitioning was used in the context of image coding for packet erasures in [90, 91]. Also note that the description indices are typically fed into separate entropy coders (one for each description) so if the description is not a power of 2, it is not an inconvenience for coding. Since these mappings are acting on the quantizer indices, they are similar to traditional channel ....

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," Proceedings SPIE - The International Society for Optical Engineering, no. 3653, 1999, pp. 998--1008.

....is insensitive to them and thus has the desired feature that the delivered quality is dependent only on the fraction of packets delivered reliably. While most of the initial work in this area has focused on the special case of descriptions [4] 5] there has been recent interest in the case also [6] [9] For example, 9] presents an efficient algorithm that transforms an MR based prioritized bitstream into a nonprioritized MD packet stream. The proposed strategy uses a purely channel coding based paradigm based on forward error correction (FEC) channel codes and exhibits a departure from ....

....individual components namely the MD FEC and the LIMD H algorithm are presented in [1] and [9] respectively wherein the two respective components were shown to significantly outperform state of the art reference systems. For example, the MD FEC algorithm was shown to outperform reference systems [6], 7] that used comparable source coders by rate savings of the order of 30 while delivering identical image qualities) In this paper, we present the performance of the end to end system as a whole especially highlighting the effectiveness of the simple interaction between the application and ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proc. SPIE Visual Communications and Image Processing Conf. (VCIP), San Jose, CA, January 1999.

....by providing di erent degrees of protection to each data classes. Similarly to the resynchronization techniques, the best results are however obtained only with a judicious packet prioritization process [18,19] In this category, layered coding [20, 21] and the Multiple Description Coding schemes [22] could be mentioned as the most promising algorithms. Optimal error resilient schemes should not only however combine techniques of above 2 HDTV stands for High De nition TeleVision 4 SUBMITTED TO IEEE TRANSACTIONS ON CSVT three categories, but also exploit the local relevance of video data. ....

Jiang W. and Ortega A. Multiple Description Coding via Polyphase Transform and Selective Quantization. In Proceedings of Visual Communications and Image Processing, VCIP'99, San Jose, USA, January 1999.

....of [23, 28, 32] Others of these use a systematic rate compatible technique [1, 2, 3, 4] which can be rate distortion optimized using the algorithms of [1, 2, 3, 4, 33, 34, 35] The present paper is an extension of these latter methods. Still others use signal processing FEC for error control [36, 37, 38, 39, 40], for which rate distortion optimization is still a research topic [41] Some papers have investigated the problem of error control using retransmission based protocols, e.g. 42, 43, 44, 45, 46, 47, 48] However, with the exception of those works listed in the previous paragraph, to our ....

W. Jiang and A. Ortega. Multiple description coding via polyphase transform and selective quantization. In Proc. Visual Communications and Image Processing, San Jose, CA, January 1999. SPIE.

....transforms are used to introduce redundancy between pairs of values in [3, 4] In [5] quantized overcomplete frame expansions are used to produce a small set of descriptions with some redundancy. Data partitioning is another technique used to add error resiliency to a source coder (e.g. [6, 7, 8, 9]) High performance compression algorithms, such as the SPIHT image coder [10] often require the source coder bits to be decoded in a sequential and uninterrupted fashion for correct interpretation. Carefully partitioning the data into separate streams prior to coding can improve the error ....

....in the higher frequency subbands. In such situations, data partitioning can be applied so that information about a specific high frequency coefficient is only sent in a subset of the packets. A similar type of data partitioning was used in the context of image coding for packet erasures in [8, 9]. The choice of mapping also has an impact on the individual entropy coders for each description. Low redundancy mappings result in high entropy descriptions, reducing the effectiveness of the entropy coders. Also, the individual descriptions may not be matched very well in terms of rate and ....

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proceedings of the SPIE, no. 3653, 1999.

....to them and thus has the desired feature that the delivered quality is dependent only on the fraction of packets delivered reliably. While most of the initial work in this area has focussed on the special case of N = 2 descriptions [4, 5] there has been recent interest in the N 2 case also [6, 7, 8, 9]. For example, 9] presents an e#cient algorithm that transforms an MR based prioritized bitstream into a non prioritized MD packet stream. The proposed strategy uses a purely channel coding based paradigm based on Forward Error Correction (FEC) channel codes and exhibits a departure from 2 There ....

....individual components namely the MD FEC and the LIMD H algorithm are presented in [9] and [23] respectively wherein the two respective components were shown to significantly outperform state of the art reference systems. For example, the MD FEC algorithm was shown to outperform reference systems [6, 7] that used comparable source coders by rate savings of the order of 30 while delivering identical image qualities) In this paper, we present the performance of the end to end system as a whole especially highlighting the e#ectiveness of the simple interaction between the application and the ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega. Multiple Description Coding via Polyphase Transform and Selective Quantization. Proceedings of the SPIE Visual Communications and Image Processing Conference (VCIP), January, San Jose, CA, 1999.

....of smaller packets. Creating multiple descriptions of a video signal tting into these small packets having the mutual re nability property is ideally matched to transmission over a network without any form of prioritization among packets. MD coding of still images has been considered recently in [4, 6, 17, 24]. Extensions to video coding have been considered in [14, 15] in the context of subband coding) and in [10, 22] in the context of predictive coding) 4.2 Predictive Multiple Description Coding with Feedback A nontrivial problem in the design of predictive MD video coders is that of encoding ....

W. Jiang and A. Ortega. Multiple Description Coding via Polyphase Transform and Selective Quantization. In Proc. SPIE Visual Commun. Image Proc. Conf. (VCIP), San Jose, CA, 1999.

....successively refine the previous blocks of data as in progressive JPEG. 4 Briefly, MD coding is concerned with the task of encoding a source into N descriptions in such a manner as to optimize the delivered source fidelity if any k out of the N descriptions are received for all k N [11] 12] [13]. The word description is synonymous with packet in this paper. 5 The R D curve for a source bitstream is a plot of the quality (equivalently distortion, i.e. lower distortion implies higher quality and vice versa) and the number of bits used to describe the source. As intuitively expected, the ....

....descent techniques and is suboptimal. Signal Processing based methods for introducing robustness into the source include the classical approaches to the MD problem that are based on quantizers [11] correlating transforms [12] and newer approaches based on wavelet polyphase decompositions [13]. Most of these approaches suffer from the issue of implementation complexity and the fact that they are difficult to generalize to more than two descriptions except for in [13] 25] Park et al. proposed the framework of an adaptive FEC (AFEC) for delay sensitive real time traffic over network ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega, "Multiple Description Coding via Polyphase Transform and Selective Quantization", Proceedings of the Visual Communications and Image Processing Conference, March, 1999.

.... [27] Along similar lines, Srinivasan and Chellappa propose a MD extension of a very high performance subband coder [15] using MD scalar quantizers [31] Jiang and Ortega propose an MD extension to the SPIHT coder of Said and Pearlman [22] by separating Zerotrees into polyphase 3 components [14]. Rogers and Cosman propose to rearrange bits at the output of one con guration of the SPIHT coder, in a way such that the loss of one packet results in an error that does not propagate beyond the image region contained in that packet [21] Mohr et al. propose the use of error correcting codes of ....

....can be explained only in terms of a DCT vs Wavelet argument. 4.2. 2 Many Descriptions As another reference, we compare the performance attained by our MD coder (now in its many packets con guration) against MD image coders of Goyal, Kova cevi c, Arean and Vetterli [12] and of Jiang and Ortega [14]. The resulting PSNRs are shown in Fig. 11. In this case we would like to point out that, using a simple form of the PET system (based on trivial repetition codes) our coder is able to attain a much wider range of tradeo s between redundancy allocations and robustness. For certain choices of ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega. Multiple Description Coding via Polyphase Transform and Selective Quantization. In Proc. SPIE Visual Commun. Image Proc. Conf. (VCIP), San Jose, CA, 1999. 29

.... often modeled through generalized multiple description coding (MDC) in which the encoder produces multiple descriptions of the source data, each exactly filling one packet [2] Recently, the generalized MDC framework has received considerable attention and many new algorithms have been proposed [2, 3, 4, 5]. In this work, we concentrate on networks which do not provide preferential treatment to packets and instead erase them randomly. Usually, however, data vary in importance. High magnitude, low frequency data that let us recognize coarse structures in an image are much more important than low ....

....packet loss rate is high enough, however, some trees may be unrecoverable. In those cases we interpolate the lowest frequency coefficients from surrounding coefficients, similar to the method in PZW [6] 3. Optimization of the Redundancy Allocation As in other unequal loss protection schemes [4, 5], we use an algorithm [7] to optimize the amounts of redundancy assigned to each wavelet coefficient tree in MD SPIHT. The allocation algorithm is based on ratedistortion tradeoffs; for a given bit rate, it minimizes 1 2 3 Descriptions sent Descriptions received 1 2 3 4 Output trees ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega. Multiple description coding via polyphase transform and selective quantization. In Proceedings of SPIE: Visual Communications and Image Processing, January 1999.

....less than N are received. This situation commonly occurs in transmissions over the Internet, where network congestion causes packet loss. The goal is to maximize the quality of the reconstruction given a set of received packets and the descriptions they contain. Other MD related papers include [2, 3, 4, 5, 6, 7]. However, many of these works do not consider the potential of forward error correction (FEC) A notable exception is the work of Puri, Ramchandran, and Kozintsev [8] We use systematic Reed Solomon (RS) codes to generate FEC. These codes are very effective at recovering erased symbols when the ....

....et al. [1] for JPEG protected by standard FEC and their frame expansion MD method. All are for 8 descriptions at about 1.25 bpp total rate, of which 20 is redundancy. Finally, in Figure 5, we compare our approach to the polyphase transform and selective quantization results of Jiang and Ortega [6], who also use the SPIHT algorithm. Each curve represents 16 descriptions at 0.5 bpp total rate, of which 20 is redundancy. Once again, MD ULP system performs significantly better and has no variability. 5. Conclusion We have presented results for the first application of explicit channel ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proceedings of Visual Communications and Image Processing, January 1999.

.... often modeled through generalized multiple description coding (MDC) in which the encoder produces multiple descriptions of the source data, each exactly filling one packet [2] Recently, the generalized MDC framework has received considerable attention and many new algorithms have been proposed [2, 3, 4, 5]. In this work, we concentrate on networks which do not provide preferential treatment to packets and instead erase them randomly. Usually, however, data vary in importance. High magnitude, low frequency data that let us recognize coarse structures in an image are much more important than low ....

....packet loss rate is high enough, however, some trees may be unrecoverable. In those cases we interpolate the lowest frequency coefficients from surrounding coefficients, similar to the method in PZW [6] 3. Optimization of the Redundancy Allocation As in other unequal loss protection schemes [4, 5], we use an algorithm [7] to optimize the amounts of redundancy assigned to each wavelet coefficient tree in MD SPIHT. The allocation algorithm is based on ratedistortion tradeoffs; for a given bit rate, it minimizes the expected distortion of the received data subject to a description loss model. ....

[Article contains additional citation context not shown here]

W. Jiang and A. Ortega. Multiple description coding via polyphase transform and selective quantization. In Proceedings of SPIE: Visual Communications and Image Processing, January 1999.

....as the Internet or wireless networks. The goal is to maximize the quality of the reconstruction given the received descriptions. Many different algorithms for MDC have been presented in the literature. Some approaches add redundancy in the form of repeated or correlated data before transmission [15, 5, 9, 3, 6, 8, 4]. In multiple description scalar and vector quantization [21, 22, 19] multiple quantizers refine each other into lower distortion quantizers if more than one description is received. All of these systems yield image quality that depends on which descriptions are received. In this work, we use ....

....descriptions are received correctly. In this case, the first three coding blocks can be decoded meaning that bytes 1 11 of the source coder output can be recovered. 1 It can be readily seen that bytes 1 11 can be recovered regardless of which two descriptions are lost. Although other systems [9, 4, 14] provide approximately balanced descriptions through various signal decompositions, MD ULP is the first system to provide perfectly balanced MDC in this manner. To assign ULP, we use the fact that a progressive source coder produces an output in which information important to image quality is ....

[Article contains additional citation context not shown here]

Wenqing Jiang and Antonio Ortega. Multiple description coding via polyphase transform and selective quantization. In Proceedings of Visual Communications and Image Processing, January 1999.

....[19] As argued, the MD constraints impose restrictions on the set of tools available to build a good coder. Yet it remains to be seen whether smarter coders within our framework, or even coders based on entirely different frameworks, are able to close this gap. In this regard, Jiang and Ortega [18] present what appears to be very promising initial results, although mostly in the low redundancy regime. Perhaps the single most important issue yet to be resolved in the context we set up in this work is whether the adaptation of MD quantizers on a per coefficient basis (instead of on a ....

....to the development of a complete system for image video transmission over the Internet. Acknowledgements. To Tanya Berger Wolf, for many useful discussions on the structure and properties of index assignments for MD quantizers. To Wenqing Jiang, for providing us a preprint of his upcoming paper [18]. ....

W. Jiang and A. Ortega. Multiple Description Coding via Polyphase Transform and Selective Quantization. In Proceedings of the SPIE Visual Communications and Image Processing Conference, San Jose, CA, 1999. To appear.

No context found.

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proceedings of VCIP'99, pp. 768-778, (San Jose, CA), Jan. 1999.

No context found.

W. Jiang, A. Ortega, "Multiple Description Coding via Polyphase Transform and Selective Quantization", Proc. of VCIP 1999.

....to one of the possible loss scenarios. In an ideal MDC channel environment, the channels are independent and data on each channel is either completely lost or received intact. This environment has been studied extensively both theoretically, 6] 7] and practically, 8] 9] 10] 11] [12], etc. The paper by Goyal [5] provides a good overview of MDC systems. Figure 1 Generic MDC system. The source is encoded in two descriptions, S 1 and S 2 . They are transmitted through the network over different channels. At the decoder, if only one description is received, the signal can be ....

W. Jiang and A. Ortega. "Multiple Description Coding via Polyphase Transform and Selective Quantization". In Proc. of Visual Communic. and Image Proc., VCIP '99, San Jose, Ca, January 1999

....transmission at different priorities and thus multiresolution techniques would not be useful, and (ii) local recovery is preferable to retransmission. These features help explain the recent revival of interest in MDC, which has led to the proposal of numerous practical MDC systems e.g. 4] [3]. Most of the recently proposed techniques are designed for memoryless coding environments and use Balanced Multiple Description Coding (BMDC) where both descriptions are coded at the same rate. One exception is the recent work of [5] where multiple description DPCM is pro posed. Here, each of ....

W. Jiang and A. Ortega, "Multiple Description Coding via Polyphase Transform and Selective Quantization", in Proc. of VCIP, 1999.

No context found.

Jiang W. and Ortega A., Multiple Description Coding via Polyphase Transform and Selective Quantization, in Proceedings of Visual Communications and Image Processing, VCIP'99, (San Jose, USA), January 1999.

No context found.

W. Jiang and A. Ortega, "Multiple description coding via polyphase transform and selective quantization," in Proc. SPIE Conf. Visual Commun. and Image Processing, San Jose, CA, 1999, vol. 3653, pp. 998-1008.

No context found.

W. Jiang, A. Ortega, "Multiple description coding via polyphase transform and selective quantization", in Visual Communications and Image Processing, SPIE 1999,vol 3653, pp. 998-1008

No context found.

Jiang W. and Ortega A., \Multiple Description Coding via Polyphase Transform and Selective Quantization," in Proceedings of Visual Communications and Image Processing, VCIP'99, (San Jose, USA), January 1999.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC