Image compression and reconstruction using pit-sigma neural networks

A high-order feedforward neural architecture, called pi(t)-sigma (pi(t)sigma) neural network, is proposed for lossy digital image compression and reconstruction problems. The pi(t)sigma network architecture is composed of an input layer, a single hidden layer, and an output layer. The hidden layer is composed of classical additive neurons, whereas the output layer is composed of translated multiplicative neurons (pi(t)-neurons). A two-stage learning algorithm is proposed to adjust the parameters of the pi(t)sigma network: first, a genetic algorithm (GA) is used to avoid premature convergence to poor local minima; in the second stage, a conjugate gradient method is used to fine-tune the solution found by GA. Experiments using the Standard Image Database and infrared satellite images show that the proposed pi(t)sigma network performs better than classical multilayer perceptron, improving the reconstruction precision (measured by the mean squared error) in about 56%, on average.