The Set Partitioning in Hierarchical Trees (SPIHT) based color image coder, commonly referred as Color SPIHT (CSPIHT) are developed to quantize and encode wavelet coefficients and have excellent rate distortion characteristic in the noise free environments. However in presence of noise they are extremely sensitive to the bit errors. It was observed that bits have different degree of vulnerability of errors. The error in some of the bits (critical bits) causes the severe degradation while errors in other bits (non-critical bits) have negligible effect in the reconstructed images. In this paper, we propose an unequal error protection (UEP) scheme, in which within each bits plane, the bits are reorganized according to their sensitivity of errors and then critical bits are protected unequally using Rate Compatible Puncturing Convolutional (RCPC) codes. The protection is reduced for higher bit-planes, by changing the puncturing rate. The simulation results show an improvement of 5-15 dB in the quality of reproduced images, as compared to the equal error protection (EEP) and unprotected CSPIHT bit-stream.
Zhou TingMing XuDongxia ChenLun Yu
Rohit Kumar MurarkaAnurag MishraKanchan Sharma