Image compression based on neural networks is presented with block classification, sub-sampling, and coding. Multilayer neural network with error back-propagation learning algorithm is used to transform the normalized image data into the compressed hidden values by reducing spatial redundancies. Image compression can basically be achieved with smaller number of hidden neurons than the numbers of input and output neurons. Additionally, the image blocks can be grouped for adaptive compression rates depending on the characteristics of the complexity of the blocks, and sub-sampling techniques can be used. The quantized output of the hidden neuron can also be entropy coded or vector quantized for an efficient transmission. Self-organizing map shows better performance than vector quantization. In computer simulation, about 25:1 compression rate was achieved using the entropy coding without much degradation of the reconstructed images, and about 40-50:1 compression rate using vector quantization or self-organizing map.
