Fast Algorithm for Optimal Error Protection of Embedded Wavelet Codes

Embedded wavelet codes are very sensitive to channel noise because a single bit error can lead to an irreversible loss of synchronization between the encoder and the decoder. Sherwood and Zeger protected a zero-tree based embedded wavelet code sent through a memoryless noisy channel by using cyclic redundancy detection codes (CRC) and channel correction codes. Chande and Farvardin proposed an optimal joint source-channel allocation strategy for such systems. We show how to accelerate their algorithm without quality loss. For grey scale test images of size 512 × 512, our speedup factors ranged from 1.2 to 6 for total bit rates between 0.25 and 1.0 bits per pixel. Moreover, by using turbo codes as channel codes, we obtained competitive peak-signal-tonoise ratio (PSNR) results.