Application of unequal error protection codes on combined source-channel coding of images

A combined source-channel encoding of images with a small amount of redundancy is investigated. The source encoder under study is based on the discrete cosine transform, adaptive quantization, and variable length encoding (VLC), while the channel encoder is based on the unequal error protection (UEP) principle. To measure the importance of each bit in the transmitted frame after the VLC, a factor of sensitivity to channel errors for each bit is defined. The factor of sensitivity is used to estimate the optimal error rate allowed for each bit in the frame that minimizes the effects of channel noise. The performance of two UEP codes constructed by the Blokh-Zyablov code construction method is evaluated for two sequences of images.<<ETX>>