A Combined Approach for Efficient Compression and Restoration of Multispectral Satellite Images

Our everyday lives are increasingly influenced by satellite images. They are an integral part of the daily news such as weather forecasting, waterbody survey, land surveying etc. Multispectral images have multiple bands, and the information will be present in at least one band. As the satellite images are very large in size, FFT2 compression technique has been proposed in this paper. This method is a lossy compression technique and granular noises get introduced in the compressed image. For restoring the images from the granular noises which are caused from sensor noises and lossy compression, an autoencoder neural network has been implemented. The neural network is trained and evaluated. The experimental results have shown that the proposed method for compression has resulted in a Compression Ratio of 11:1 and an autoencoder neural network model removes the granular noise with an efficiency of 83%