Our further research is based on several assumptions. First, we concentrate on considering lossy compression of single-channel images or componentwise compression of multichannel (hyperspectral) data. Clearly, joint [three-dimensional (3-D)] compression of hyperspectral data is able to provide better CR,5,7,21,27 but it is necessary to understand from the beginning how to set the PCC for each (sub-band) image. Second, a simple case of additive white Gaussian noise is considered. Here, one has to keep in mind that if noise is signal dependent (Poissonian, additive and Poissonian, additive and multiplicative, etc.), there is a VST able to convert such noise to the additive one.22,27,30 This is possible if parameters of signal-dependent noise are estimated with appropriate accuracy and modern existing methods.23,24 Third, major attention is paid to consider DCT-based coders for which it is simpler to provide compression in an OOP neighborhood17,18,20,21 and for which we are more experienced in their design and performance analysis. Finally, we will focus on analyzing two metrics: conventional PSNR and visual quality metric PSNR-human visual system (HVS)-masking (M).31 It might seem that these metrics (especially the latter one) do not have a direct relation to RS data processing. However, one has to take into account the following. On one hand, the criteria used in RS data compression are application oriented.8,32 On the other hand, metrics exploited in compression of conventional images are also used in lossy compression of hyperspectral images and this relates to HVS metrics as well.33 One reason is that HVS metrics highly correlate with such important properties as edge, detail, and texture feature preservation by lossy compression techniques.