Image Transmission Based on DnCNN Adaptive Selection of Significant Bit Optimization

To address the problem of low reliability in soft decoding for wireless image transmission in extremely low signal-to-noise ratio environments, a method based on Deep Convolutional Neural Network combined with an adaptive highposition bit selection algorithm(DnCNN-AHPBS) is proposed to enhance quality of image transmission in low signal-to-noise ratio environments. Firstly, the image information is mapped to a bitstream for 16 QAM modulation, and then the modulation information is encoded by 5G new radio access technology LDPC channel coding, and then transmitted wirelessly through the MIMO-OFDM communication system. Secondly, after being wirelessly transmitted in a low signal-to-noise ratio environment, the bits are restored to the image structure in DnCNN assisted optimization. Finally, the high position bit information selected by the DnCNN-AHPBS algorithm is mapped into a constellation diagram for flipping and replacement. This method adjusts the initial log-likelihood ratio information of the Minimum Offset Sum (OMS) algorithm to obtain an optimized log likelihood ratio and reduce the error rate of image transmission in low signal-to-noise ratio environments. The simulation results indicate that a specific bit error rate will result in a 1.4 dB bit signal-to-noise ratio gain at extremely low signal-to-noise ratios. The DnCNN-AHPBS algorithm has to some extent improved the reliability of wireless transmission soft decoding in extremely low signal-to-noise ratio environments.