Lossless compression of hyperspectral images using differential pulse code modulation

Aerial and satellite images have often high resolution and they cover large areas. As a result the dimensions of the images are large and require huge amounts of storage resources. Lossless compression is one technique used to alleviate this problem. Hence a lossless compression algorithm for hyperspectral images is proposed. The spectral (interband) correlation is much stronger than the spatial (intraband) correlation in hyperspectral images, and it must be utilized for maximal compression performance. The proposed algorithm makes use of the spectral correlation existing between successive bands of hyperspectral images. After the spectral prediction, the spatial prediction of each band is done if the intraband correlation of the particular band is greater than the threshold value. Arithmetic encoding technique is used for encoding the obtained residues and a compressed image is obtained and it is decoded, which is converted to a 2-d array and the operation performed at the earlier stage is reversed to get back the decompressed image. The compression ratio for hyperspectral images and the features of the algorithm are compared to those of other lossless image compression algorithms and a satisfactory ratio is obtained.