Error concealment using multiple description coding and LIoyd-max quantization

Abstract

There has been increasing usage of Multiple Description Coding (MDC) for error concealment in non ideal channels. This paper attempts to conceal the error and reconstruct the lost descriptions by combining MDC and LIoyd-max quantizer. At first original image downsampled to four subimages then by using wavelet transform each subimage is mapped to transform domain then descriptions are quantized by LIoyd-max and coded. Since in proposed method wavelet transform is been used, there could be no blocking effect as compared to DCT transform. The results show that average MSE (mean square error) for our proposed method in comparison with DCT method (in other paper) has decreased from 240 to 152 in 0.625 bpp (bit per pixel), from 161 to 107 in 1 bpp and from 96 to 73 in 2 bpp in rate-distortion performance. Therefore, this method is suitable for low capacity channels. By losing descriptions, the obtained image is still in good quality (subjective evaluation and PSNR values) as compared to a method which is DCT and MDSQ (multiple description scalar quantization).