This work presents a novel recurrent neural network called the recurrent Pi-sigma neural network. The proposed network has been used as 2-D predictor in differential pulse code modulation (DPCM) image coding. The advantage of this type of architecture is that it explores both the multi-linear interactions between the input pixels as well as the temporal dynamics of the image formation process. The network was trained using the dynamic backpropagation algorithm. Fifteen images were used to test the performance of the network. Extensive simulation results have shown an average peak signal to noise ratio (PSNR) of 26.2 dB at a transmission rate of 1 bit/pixel.

The recurrent Pi-sigma neural network effectively predicts 2-D image coding with an average peak signal to noise ratio of 26.2 dB at 1 bit/pixel.

A new recurrent polynomial neural network for predictive image coding

Key Takeaway: The recurrent Pi-sigma neural network effectively predicts 2-D image coding with an average peak signal to noise ratio of 26.2 dB at 1 bit/pixel.