Compression of Spectral Images
Arto Kaarna · 2007
In this section we have collected experiences when different spectral images were compressed in a lossy manner with various methods described in the previous sections. The following abbreviations are used: • CL-W: wavelet transform in the spectral reduction followed by clustering, • CL-P: PCA in the spectral reduction followed by clustering, • WT-3M: the three-dimensional wavelet transform, Chui-Lian multiwavelets (Chui & Lian, 1996), • WT-3H: the three-dimensional wavelet transform, Haar wavelet, • SP-P: PCA in the spectral reduction and SPIHT (Said & Pearlman, 1996) in the spatial dimensions, • JP2K-P: PCA in the spectral reduction and JPEG2000 (Taubman & Marcellin, 2002) in the spatial dimensions, • JPG-P: PCA in the spectral reduction and DCT/JPEG (Rabbani & Jones, 1991) in the spatial dimensions, • SP-O: SPIHT in the spatial dimensions, no spectral reduction, • JPG-O: DCT/JPEG in the spatial dimensions, no spectral reduction, • JP2K-O: JPEG2000 in the spatial dimensions, no spectral reduction. In earlier experiments (Kaarna et al., 2000), it was found, that PCA and wavelets performed best with clustering, and, thus, from the comparisons we leave out the other possible variations. The last three methods provided trivial solutions to the compression of spectral images. These methods applied SPIHT, DCT/JPEG, or JPEG2000 to the spatial dimensions of the images without any compression in the spectral dimension. Thus, we could get some