Image Compression Using Adaptive Neural Network Based on Pixels Information

The aim of this work is to create an image compression technique using one hidden layer feed forward neural network in which the number of the input neurons is determined adaptively according to pixels values. The technique proposes scanning the image pixels values and takes one sample value from the occurring values ranging from 0 to 255 gray scales. The taken values are fed into the neural network as input signals instead of the original image pixels. This process exhibits decreasing the number of the input neurons, the computing size, memory space and better rate of convergence of the back propagation training algorithm. The proposed work has been demonstrated through several experiments and shows very promising results in compression rates as well as in the reconstructed image