Increasing spectral resolution of hyper spectral images while decreasing spectral variability using deep generative model

Human vision is known to extend between 400 and 700nm, in other words, it can only see visible light in three bands which are red, green and blue, however in scientific inquiry, it is necessary to look far beyond than that, into the infrared ranges whereas it can divides the spectrum into several further bands, for this reason a recent advances in optical analysis introduced a technology called hyper spectral imaging that allows for the detection and identification of objects from a large scale of images providing enormous number of wavebands, such imaging is based on the idea that it can record underlying processes (e.g., chemical traits, biophysical qualities, etc.) down to the pixel level, In other words it is based on the idea of a continuous spectrum, or of narrow, contiguous spectral bands, which is crucial and allows for the greatest possible use of the information. Thus, by examining their tiny spectrum features, it is able to recognize surface objects (or atmosphere gases/particles) and evaluate their optical qualities. However, to attain maximum result's efficiency hyperspectral imagery must overcome numerous conditions and circumstances that reduce the capability to interpret and analyze the data which can affect its performance leading to weak spectral Resolution and high spectral variability. These conditions can be as atmospheric changes or in some cases, it can occur a lot of materials in the same spectral signature as it gets more difficult to correctly define and measure the basic substances.