Hyperspectral Data Compression by Using Rational Function Curve Fitting in Spectral Signature Subintervals

摘要

Hyperspectral data is a collection of many images, all of which are from a single terrestrial landscape, but at different and adjacent wavelengths (bands), and often in whole or in part wavelengths of 400 to 2500 nm. For each pixel of hyperspectral data, The curve obtained by plotting brightness intensities in different bands in terms of band numbers is known as the spectral signature or the spectral reflection curve (SRC).Compression methods are based on transformations coding such as discrete cosine transform (DCT), discrete wavelet transform (DWT), or principal component analysis (PCA) are one of the most effective ways to eliminate image correlations and to reduce their volume. But all of these methods suffer from a common mistake, which is that they do not consider spectral reflectance curve Algebraic - Geometric features and a rich source of information are neglected as sequence of primary features. Another method is based on curve fitting which is used due to its effect on image spectrum exclusively in Compression Hyperspectral Images. This method uses the Spectral Signature Image to reduce the feature. This method has possessed very good results compared with previous methods such as PCA, but in compression by using this method, the SRC approximated curve in some points has distortion. In this paper, we tried to use a specific way finding distortion points and interval SRC to non-overlapping adjacent intervals in order to resolve this distortion. Using the proposed method, in addition to eliminating distortion, the PSNR level has much increased and the reconstructed image quality is very similar to the original image.