Exploiting the UEP property of polar codes to reduce image distortions induced by transmission errors

This paper considers the application of polar codes in image transmissions. When the polar codeword length is finite, the information bits in different positions experience different but predictable unequal post-decoding error probabilities. Exploiting this property, we propose to order the information bits in a polar codeword according to their erroneously transmission induced image distortions. Specifically, the most significant bits (MSB, left-most bits) of each pixel in an image are transmitted through the lower error probability bit-channels (logical channel post the polar code) and the Least Significant Bits (LSB, right-most bits) of each pixel are transmitted through higher error probability bit-channels. In this way, the Polar codes output bit streams with different importance will be given unequal error protection (UEP), resulting in smaller distortion to the image. Our analysis and simulation results reveal that the image transmission system coded by polar codes with our UEP scheme yields more than 7dB gain over the same polar codes image transmission system without UEP, when the delivered image Peak Signal to Noise Ratio (PSNR) is near 40dB.