Digital Image Transmission with Polar Codes and Median Filtering
Kuen‐Tsair Lay, Han-Ting Huang · 2019
In digital images, different bits can carry different degrees of importance. Therefore, unequal error protection (UEP) is quite often incorporated into the transmission of digital images. The so-called "protection' is achieved via error control coding (ECC), for which the most investigated ECC class in recent years - namely, polar codes - are adopted in this work. One key issue with polar codes is code construction - determination of frozen bit-channels and information bit-channels. Moreover, for UEP, the order of the information bit-channels according to which the various bits in image pixels are to be placed needs to be determined. One specific ordering scheme is proposed in this paper and compared to an existing method. One phenomenon observed in polar decoding is that whenever there occurs an erroneously decoded bit, the succeeding bits are also likely to become erroneously decoded. In terms of visual quality, this problem translates into continous line-segments of obviously wrong lightness. To counter this problem, the technique of median filtering combined with pixel-interleaving is adopted in this work. It turns out that our scheme produces good results, in terms of the PSNR (peak signal-to-noise ratio) and the visua quality of the recovered images.