Fig. 1. The MA-FGS architecture.

Fig. 2. The PSNR performance of encoding the test video sequences at different bit rates. (A) QCIF Foreman with 10 fps and 32 kbps for the base layer. (B) QCIF Akiyo with 10 fps and 16 kbps for the base layer.

Fig. 3. The adopted UEP structure.

Fig. 4. The R–D relationship within each bitplane of the third frame of the QCIF Foreman video sequence at 10 fps and 32 kbps for the base layer. The previous two enhancement frames are coded at 6400 bits. The R–D curves for all the bitplanes are placed in an increasing order (starting with the first bitplane) from left to right and from top to bottom. The x-axis represents number of bits while the y-axis represents MSE.

Fig. 5. The results of the channel distortion estimation for the QCIF Foreman video sequence coded at 10 fps and 96 kbps with 32 kbps for the base layer, no FEC and 5% packet loss rate which only occurs in the latter 30% part of each frame. (A) Δ = 1. (B) Δ = 5. (C) Δ = 10.

Fig. 6. The PSNR performance of transmitting the video sequences with different total bandwidth over a BSC channel with  = 0.1. (A) QCIF Foreman. (B) QCIF Akiyo.

Fig. 7. The PSNR performance of transmitting the test video sequences with a fixed total bandwidth over BSC channels with different SERs. (A) QCIF Foreman with 128 kbps. (B) QCIF Akiyo with 64 kbps.

Fig. 8. The average PSNR performance of all the video frames for transmitting the test video sequences with a fixed total bandwidth over a BSC channel with  = 0.1. (A) QCIF Foreman with 128 kbps. (B) QCIF Akiyo with 64 kbps.

Fig. 9. An example of the channel distortion and the corresponding coding mode for each frame over one simulation in the case of transmitting the QCIF Foreman at 128 kbps over a BSC channel with  = 0.1. (A) Channel distortion. (B) Coding mode.

The symbol definitions