Rate-distortion optimization for the SPIHT encoder

Clustering source output bits and equalizing bit error sensitivity to improve the quality and robustness of transmitted images over wireless channels

“…According to the bitplane positions or the estimated BES distributions [21], they estimate the BES of sign bits and the BES of magnitude bits, respectively [15]. However, their estimation is taken in the wavelet-subband domain rather than in the image domain, which incurs some distortions [12]. Furthermore, the BES of bits in LISS of C-SPIHT are difficult to be estimated.…”

Fast rate-distortion estimation and optimization for wavelet video coding

“…As Figure 1 illustrates, there is significant high frequency content which will cause the significance mapping algorithm just described to output many bits until the tree descent reaches the significant coefficients. Lin and Gray provide numerical evidence for this effect in [4].…”

“…The solution, presented in [4], is to divide the spatial domain image into tiles, and subject each to independent SPIHT encoding at different significance levels. The operating principle behind this is the strong significant coefficient clustering in the spatial domain.…”