Noise Reduction and Concealment for Block Encoded Images

A simple scheme to detect the presence of the channel noise on the vector quantized images are proposed. The channel noise effect is removed or concealed by utilizing the continuity of the image signal and the availability of the complete codebook at the receiver. Several coding techniques (such as channel-optimized vector quantization and vector quantization with pseudo-Gray coding) are tested. It is found that the proposed technique always improves the image quality under noisy channel. In addition, it provides the best image quality when vector quantization with pseudo-Gray coding is used as the coding technique. Under this scheme, the proposed technique provides 0.2-1.2 dB performance gain (in average) for a binary symmetric channel with bit error rate 0.1-1%, respectively. >