#### DMCA

## Fair Rate Allocation of Scalable Multiple Description Video for Many Clients (2005)

Venue: | In Proc. of Visual Communications and Image Processing |

Citations: | 8 - 3 self |

### Citations

562 | Splitstream: High-bandwidth multicast in a cooperative environment,”
- Castro
- 2003
(Show Context)
Citation Context ...d by Rn. Hence, each successive layer i 6 7 Mshas a rate Ri − Ri−1. The rate RD of each of the M descriptions is then given by: where R0 = 0, and αl = 1 l(l + 1) RD = RD = M� l=1 Rl − Rl−1 l M� αlRl, =-=(2)-=- l=1 for l = 1, 2, M − 1, and αM = 1/M. (3) A node or client in the P2P network with bandwidth Rc receives m = ⌊ Rc ⌋ descriptions. After decoding these m RD descriptions, the client experiences a com... |

302 |
Design of Multiple Description Scalar Quantizers
- Vaishampayan
- 1993
(Show Context)
Citation Context ...RD and M are fixed, the maximization function becomes: where, {R1, R2 . . . , RM } = arg max R1,R2...,RM M� � bj+1 j=0 bj bi = i RD for i = 0, 1, . . . , M bM+1 = ∞. fRc(r)F C(r; R1, R2 . . . , Rj)dr =-=(7)-=-s4.2. Minimal MSE Fairness Criterion The first – and the most straightforward – criterion we consider is to average the distortion D(R1, R2 . . . , Rm) = D(R1, Rm) over all clients: ˆD(R1, R2 . . . , ... |

178 | Resilient peer-to-peer streaming
- Padmanabhan, Wang, et al.
- 2003
(Show Context)
Citation Context ...D, this criterion can be solved numerically using the Lagrange multiplier method, as discussed in the work of Puri et al. 4 L(R1, R2 . . . , RM , λ) = ⎛ M� M� Cj D(R1, Rj) + λ ⎝ j=0 j=1 αjRj − RD ⎞ ⎠ =-=(10)-=- After equating the partial derivatives to zero, we obtain a set of equations for which the roots can be found numerically: 1 M� α1 j=1 ∂D(R1, Rj) Cj + λ = 0 (11) ∂R1 Ci αi ∂D(R1, Ri) + λ ∂Ri = 0 for ... |

176 | Chainsaw: Eliminating trees from overlay multicast
- Pai, Kumar, et al.
- 2005
(Show Context)
Citation Context ...he partial derivatives to zero, we obtain a set of equations for which the roots can be found numerically: 1 M� α1 j=1 ∂D(R1, Rj) Cj + λ = 0 (11) ∂R1 Ci αi ∂D(R1, Ri) + λ ∂Ri = 0 for i = 2, . . . , M =-=(12)-=- M� αlRl − RD = 0 (13) l=1 In Ref. 4, a method is presented to solve these simultaneous equations such that after optimization Ri < Ri+1 holds for all rates. Unfortunately, the required conditions can... |

150 | On multiple description streaming with content delivery networks
- Apostolopoulos, Wong, et al.
- 2002
(Show Context)
Citation Context ...D(R1, Rj) + λ ⎝ j=0 j=1 αjRj − RD ⎞ ⎠ (10) After equating the partial derivatives to zero, we obtain a set of equations for which the roots can be found numerically: 1 M� α1 j=1 ∂D(R1, Rj) Cj + λ = 0 =-=(11)-=- ∂R1 Ci αi ∂D(R1, Ri) + λ ∂Ri = 0 for i = 2, . . . , M (12) M� αlRl − RD = 0 (13) l=1 In Ref. 4, a method is presented to solve these simultaneous equations such that after optimization Ri < Ri+1 hold... |

123 | A peer-to-peer architecture for media streaming, in
- Tran, Hua, et al.
- 2004
(Show Context)
Citation Context ... 2 D(R1,Rj) . We can optimize the above criterion in a similar way as the minimum MSE criterion. The Lagrangian function is then given by ⎛ ⎞ M� M� L(R1, R2 . . . , RM , λ) = CjPSNR(R1, Rj) + λ ⎝ ⎠ . =-=(15)-=- After partial differentiation to Ri we obtain the following set of equations: 1 M� Cj α1 j=1 Ci αi j=0 j=0 αjRj − RD 1 ∂D(R1, Rj) + λ = 0 (16) D(R1, Rj) ∂R1 1 ∂D(R1, Ri) + λ D(R1, Ri) ∂Ri = 0 for i =... |

100 | P2Cast: Peer-to-Peer Patching Scheme for VoD Service
- Guo, Suh, et al.
- 2003
(Show Context)
Citation Context ... to zero, we obtain a set of equations for which the roots can be found numerically: 1 M� α1 j=1 ∂D(R1, Rj) Cj + λ = 0 (11) ∂R1 Ci αi ∂D(R1, Ri) + λ ∂Ri = 0 for i = 2, . . . , M (12) M� αlRl − RD = 0 =-=(13)-=- l=1 In Ref. 4, a method is presented to solve these simultaneous equations such that after optimization Ri < Ri+1 holds for all rates. Unfortunately, the required conditions can only be verified for ... |

98 |
Chaining: A generalized batching technique for video-on-demand systems
- Sheu, Hua, et al.
- 1997
(Show Context)
Citation Context ...for Gaussian i.i.d. sources (a0 = 0) ,but also models the enhanced coding efficiency for autocorrelated sources such as video. For R → ∞, the slope of the curve becomes the well-known 50 45 40 35 (b) =-=(1)-=- rs6dB per bit. For smaller R, the curve has a larger slope, which slowly decays to the 6dB bound. The top curve in Figure 4(b) shows the resulting model. We see in Figure 4(a), that the slope of the ... |

83 |
P2VoD: Providing fault tolerant video-on-demand streaming in peer-to-peer environment
- Do, Hua, et al.
- 2004
(Show Context)
Citation Context ... measure often used in video compression, namely peak-SNR. When we average the PSNR over all clients, the following criterion is obtained: �PSNR(R1, R2 . . . , RM ) = bj+1 M� � fRc(r) PSNR(R1, Rj) dr =-=(14)-=- j=0 bj 255 where PSNR(R1, Rj) = 10 log10 2 D(R1,Rj) . We can optimize the above criterion in a similar way as the minimum MSE criterion. The Lagrangian function is then given by ⎛ ⎞ M� M� L(R1, R2 . ... |

80 | Generalized multiple description coding with correlating transforms
- Goyal, Kovacevic
- 2001
(Show Context)
Citation Context ...e most straightforward – criterion we consider is to average the distortion D(R1, R2 . . . , Rm) = D(R1, Rm) over all clients: ˆD(R1, R2 . . . , RM ) = bj+1 M� � M� fRc(r)D(R1, Rj) dr = Cj D(R1, Rj), =-=(8)-=- j=0 bj where the number of clients receiving i + 1 out of M descriptions is Ci, computed as: Ci = j=0 bi+1 � fRc(r) dr for i = 0, 1, . . . , M. (9) bi Given a fixed number of description M and a fixe... |

65 |
Multiple Description Source Coding through Forward Error Correction Codes
- Puri, Ramchandran
- 1999
(Show Context)
Citation Context ...istortion model In order to be able to find an optimal rate allocation, we need to have an analytic model of the RD curves. First we model the single layer curve. D(R) = σx 2 2 −2(R+a0(1−2−b 0 R )) . =-=(4)-=- This model embeds the information-theoretical bound for Gaussian i.i.d. sources (a0 = 0) ,but also models the enhanced coding efficiency for autocorrelated sources such as video. For R → ∞, the slope... |

47 | A measurement study of the BitTorrent peer-to-peer file-sharing system
- Pouwelse, Garbacki, et al.
- 2004
(Show Context)
Citation Context ...d the fairness criterion, the optimal compression parameters M, R1, . . . , RM can then be found by maximizing {M, R1, R2 . . . , RM } = arg max M,R1,R2...,RM � 0 ∞ fRc(r)F C(r; R1, R2 . . . , RM )dr =-=(6)-=- In most cases the rate allocation problem (finding optimal values for R1, R2 . . . , RM given RD and M) can only be solved numerically. The values of RD and M, however, can either be optimized numeri... |

21 | Error-resilient video compression through the use of multiple states
- Apostolopoulos
(Show Context)
Citation Context ...) = bj+1 M� � M� fRc(r)D(R1, Rj) dr = Cj D(R1, Rj), (8) j=0 bj where the number of clients receiving i + 1 out of M descriptions is Ci, computed as: Ci = j=0 bi+1 � fRc(r) dr for i = 0, 1, . . . , M. =-=(9)-=- bi Given a fixed number of description M and a fixed rate per description RD, this criterion can be solved numerically using the Lagrange multiplier method, as discussed in the work of Puri et al. 4 ... |

17 | Real-time video delivery using peer-topeer bartering networks and multiple description coding
- Pouwelse, Taal, et al.
- 2004
(Show Context)
Citation Context ... Mshas a rate Ri − Ri−1. The rate RD of each of the M descriptions is then given by: where R0 = 0, and αl = 1 l(l + 1) RD = RD = M� l=1 Rl − Rl−1 l M� αlRl, (2) l=1 for l = 1, 2, M − 1, and αM = 1/M. =-=(3)-=- A node or client in the P2P network with bandwidth Rc receives m = ⌊ Rc ⌋ descriptions. After decoding these m RD descriptions, the client experiences a compression distortion of D(R1, R2 . . . , Rm)... |

6 | Scalable multiple description coding for video distribution
- Taal, Pouwelse, et al.
- 2004
(Show Context)
Citation Context ...R1)+a(R1)(1−2−b(R 1 )(Rm−R 1 ) )) Functions a(R1) and b(R1) are fit to the experimentally observed rate-distortion behavior of the coder. 4.1. Rate Allocation Problem 4. FAIR MDC STREAMING for m ≤ M. =-=(5)-=- If we have to serve a large number of different clients, all with a different bandwidth Rc, we have to trade-off quality and redundancy of the descriptions. Making an optimal trade-off is not trivial... |

2 |
Chaining: A generalized batching technique for video-on-demand systems
- Tavanapong
- 1997
(Show Context)
Citation Context ...m SPIE Digital Library on 21 May 2010 to 131.180.130.114. Terms of Use:shttp://spiedl.org/terms has a rate Ri −Ri−1. The rate RD of each of the M descriptions is then given by: RD = M∑ l=1 Rl −Rl−1 l =-=(1)-=- RD = M∑ l=1 αlRl, (2) where R0 = 0, and αl = 1 l(l + 1) for l = 1, 2,M − 1, and αM = 1/M. (3) A node or client in the P2P network with bandwidth Rc receives m = RcRD descriptions. After decoding ... |