### TABLE III UNEQUAL ERROR PROTECTION AND EQUAL ERROR PROTECTION CHANNEL AND SOURCE RATE DISTRIBUTION FOR A FIXED

2004

Cited by 3

### TABLE I SIMULA a203 TION PARAMETERS FOR TRANSMISSION OF PTVC CODED VIDEO SEQUENCES OVER INTERNET ERASURE CHANNELS USING UNEQUAL ERASURE PROTECTION ACCORDING TO COMMON TEST CONDITIONS

2001

Cited by 13

### Table 2(a). Average SNR (dB) values at BER = 0.001 for images coded by VSPIHT (dim 16) and unequal block codes.

"... In PAGE 4: ... The SNR values obtained for a total of 50 runs for each case is averaged to obtain a final representative SNR for the image at that bitrate and BER. The results thus obtained are tabulated in Table2 . In the absence of noise, the SPIHT coder without arithmetic coding perform a little better than the VSPIHT(16) coder also without arithmetic coding.... In PAGE 4: ... For the non-critical bits, the same BCH code is used for the first three passes, a (31, 26) 1-error correcting BCH code is used for the fourth and fifth passes, a (63, 57) 1-error correcting BCH code is used for the sixth pass, and the remaining passes are left unprotected. The SNR results obtained on an average by VSPIHT of dim 16 in conjunction with this error protection scheme, from several runs at various bit rates, is presented in Table2 (a). In the second block code implementation for a BER of 0.... In PAGE 4: ... For the non-critical bits, the same Golay code is used for the first three passes, a (31, 21) 2-error correcting BCH code is used for the next three passes, and the (15, 11) BCH code is used for the remaining passes. The SNR results are presented in Table2 (b). In the first convolution code based implementation for a BER of 0.... In PAGE 4: ... The non-critical bits from the first three passes are protected with the same convo- lution code but with a block size of 180 information bits (279 channel bits with zero termination). Those from the remain- ing passes are coded with a (4, 3, 2) convolution code with a block size of 180 information bits (248 channel bits with zero Table2 . Average SNR values for various bitrates and various bit-error- rates for the Lena Image coded by SPIHT and VSPIHT (dim 16).... In PAGE 4: ... Table2 (b). Average SNR (dB) values at BER = 0.... In PAGE 5: ...conjunction with the above unequal protection scheme for vari- ous images are presented in Table2 (c). In the second convolu- tion code implementation at a BER of 0.... In PAGE 5: ... Those from the remaining passes are protected with a (2, 1, 8) convolution code with a block size of 180 information bits (376 bits with zero termination). The aver- age SNR results for VSPIHT (dim 16) in conjunction with the above unequal protection scheme for various images are pre- sented in Table2 (d).All the convolution codes used are taken from the tables in [10].... In PAGE 5: ... The critical bits and the non-critical bits from the first pass are protected with a (63, 18) 10-error correct- ing BCH code with interleaving degree 10 (effective block length 630 channel bits). The non-critical bits from the second and third passes are protected with a (31, 11) 5-error correcting BCH code with interleaving degree 20 (effective block length 620 channel bits), those from the fourth pass are protected with a (63, 30) 6-error correcting code with interleaving degree 10 (effective block length 630 channel bits), those from the fifth pass are protected with a (63, 39) 4-error correcting BCH code with interleaving degree 10 (effective block length 630 channel bits), those from the sixth pass are protected with a (63, 45) 3- Table2 (c). Average SNR (dB) values at BER = 0.... In PAGE 5: ...78 29.81 Table2 (d). Average SNR (dB) values at BER = 0.... ..."

### Table 3: Optimal unequal error protection architecture for RCPT and RCPC

"... In PAGE 4: ... Finally, the puncturing architecture of the bitstream is derived so that the final source-channel coded bitstream equals 20 kbps for the 4-PSK RCPC or 30 kbps for the 8 PSK RCPT. Table3 summarizes the puncturing architecture for both channel encoders and for source coding bitrates in steps of 2 kbps. 4.... ..."

### Table 1. Comparison of equal and unequal error protection on a frame basis.

"... In PAGE 3: ... With the models only a60a150a146a151a60 a90a114a148 a8 a60 a47 a107 a75 a60 a12 compression and trans- mission simulations are needed. In Table1 the near-optimal rate allocation found by this procedure is compared with equal error protec- tion for a 6-frame sequence. The overall rate is set to a152 a49 a78 a49a17a49a17a49 bits and three source rate allocations are illus- trated.... ..."

### Table 3: Optimal unequal error protection architecture for AWGN and Rayleigh fading channels (max.10 kbauds).

### Table 2(c). Average SNR (dB) values at BER = 0.01 for images coded by VSPIHT (dim 16) and unequal convolution codes.

"... In PAGE 4: ... The SNR values obtained for a total of 50 runs for each case is averaged to obtain a final representative SNR for the image at that bitrate and BER. The results thus obtained are tabulated in Table2 . In the absence of noise, the SPIHT coder without arithmetic coding perform a little better than the VSPIHT(16) coder also without arithmetic coding.... In PAGE 4: ... For the non-critical bits, the same BCH code is used for the first three passes, a (31, 26) 1-error correcting BCH code is used for the fourth and fifth passes, a (63, 57) 1-error correcting BCH code is used for the sixth pass, and the remaining passes are left unprotected. The SNR results obtained on an average by VSPIHT of dim 16 in conjunction with this error protection scheme, from several runs at various bit rates, is presented in Table2 (a). In the second block code implementation for a BER of 0.... In PAGE 4: ... For the non-critical bits, the same Golay code is used for the first three passes, a (31, 21) 2-error correcting BCH code is used for the next three passes, and the (15, 11) BCH code is used for the remaining passes. The SNR results are presented in Table2 (b). In the first convolution code based implementation for a BER of 0.... In PAGE 4: ... The non-critical bits from the first three passes are protected with the same convo- lution code but with a block size of 180 information bits (279 channel bits with zero termination). Those from the remain- ing passes are coded with a (4, 3, 2) convolution code with a block size of 180 information bits (248 channel bits with zero Table2 . Average SNR values for various bitrates and various bit-error- rates for the Lena Image coded by SPIHT and VSPIHT (dim 16).... In PAGE 4: ...21 32.99 Table2 (a). Average SNR (dB) values at BER = 0.... In PAGE 4: ...51 30.74 Table2 (b). Average SNR (dB) values at BER = 0.... In PAGE 5: ...conjunction with the above unequal protection scheme for vari- ous images are presented in Table2 (c). In the second convolu- tion code implementation at a BER of 0.... In PAGE 5: ... Those from the remaining passes are protected with a (2, 1, 8) convolution code with a block size of 180 information bits (376 bits with zero termination). The aver- age SNR results for VSPIHT (dim 16) in conjunction with the above unequal protection scheme for various images are pre- sented in Table2 (d).All the convolution codes used are taken from the tables in [10].... In PAGE 5: ... Table2 (d). Average SNR (dB) values at BER = 0.... ..."

### TABLE I RS codes for unequal error protection

### Table 3. Average SNR (dB) values over a dual-state channel with BER = 0.01 (95% of the time), and = 0.1 (5% of the time) with average burst length 100 bits, for images coded by VSPIHT (dim 16) and unequal interleaved block codes.

"... In PAGE 5: ... The codes and the inter- leavers are so chosen that the effective channel block length never exceeds 630 bits, thus preserving sufficiently fine granularity of embedding. The SNR results are shown in Table3 . An equal protection scheme (like [7]), applied on an inherently vulnerable bitstream as that obtained from arith- metic-coded scalar SPIHT [2], needs to designed for the worst-case scenario, and therefore loses efficiency.... ..."