Satellite image compression using a bottleneck network

The main objective of this paper is to present a simple neural network suitable for a big sized image compression like a satellite image. The defined network of NxMxN neurons represents the input, hidden, and output layers respectively. Since M << N, then this type of structure is referred to as a bottleneck neural network. A new mechanism of training method and robustness network is satisfied, when this network is trained to images having small size, and tested to satellite images. The input image is segmented into L blocks each has N elements. Each sub-image presented to the network as input, would appear nearly the same as the output layer. Many experiments were done for different satellite images and the goodness of fit between the original image and the reconstructed was found to be about nearly 95% with compression ratio 4.2:1 even for new images that the network did not learn about. It was found that the network is not affected by the geometrical distortions like translation, size, and rotation.