Decorrelation of IRS LISS-III data for lossless compression

The present paper analyses the scope of decorrelation as a tool for IRS data compression. Such a study was earlier carried out for Thematic Mapper. Now this study for 4 band data of IRS established the validity of spectral decorrelation for lossless compression. Most lossless compression algorithms in remote sensing applications use only spatial decorrelation. The authors concentrate mainly on spectral decorrelation. They first investigate the potential and limitations of spectral decorrelation methods for study of IRS data. Then spectral decorrelation usage is considered to achieve better compression ratio. The statistical properties of IRS data are also investigated. The image corresponds to agricultural area and forest areas. The statistical information indicates that some bands such as bands 1, 2 and 3 have strong spectral correlation while some other bands such as 1 and 4 have weak spectral correlation. This suggests that, by using spectral decorrelation, the authors may obtain better results for bands 1, 2 and 3. Experiments are carried out using this test IRS image. The image consists of four spectral channels and a spatial dimension of 1K×1K pixels. In this paper, lossless compression techniques proposed for Landsat TM multispectral images are applied for IRS data. These techniques provide methods to efficiently compute the optimal band combination and band ordering based on the statistical properties of multispectral image data. The authors´ study on IRS LISS-III data showed that these techniques are capable of achieving higher compression ratios