Data compression for image recognition

This paper describes the application of an artificial neural network (ANN) to different features of compression and decompression. ANNs have a simple and efficient method for data compression/decompression and in certain cases prove to be advantageous over coding techniques such as Huffman and binary. Two architectures of neural networks were built using the backpropagation method. Suggestions on how to increase the performance of the network have also been outlined. The well proven backpropagation network was studied and experimented in detail. Parameters such as network training time, weights on the connections, accuracy, noise level, number of neurons in the hidden layer etc. were studied and discussed to obtain estimates of the effect each parameter has on the network. Apart from the training stage for the neural network, the network was able to compress/decompress digits fairly efficiently and correctly.