perClass Mira provides an interactive visualization tool enabling the user to define custom spectral indices highlighting different aspects of spectral data. By a spectral index, we mean a quantity computed from spectral information at a pixel level.


For example, one commonly-used spectral index is NDVI which, in broad terms, highlights a difference between visible and near-infrared reflectance. It is used in remote sensing in order to estimate vegetation coverage NDVI = (NIR - RED) / (NIR + RED). In the NDVI equation, the RED and NIR terms corresponds to integrated reflectance in red (visible) and near-infrared areas of the spectrum.


In perClass Mira, Visualization or spectral feature extraction tools allow the user to define their own spectral indices either based on wavelength specification or interactively. While the former lets us work with spectral indices defined in literature, the later provides a powerful way of discovering hidden signals in spectral data by interactive experimentation with a direct visual feedback.


In order to use the interactive visualization, we organize the perClass Mira window in the following way: we position the Visualization panel above the Spectra.



We may select the type of spectral feature (generic equation) in the combo box located in the Visualization panel.



We select the (A-B)/(A+B) equation covering the NDVI type of index discussed above, this creates a new spectral feature called "F1:(A-B)/(A+B)".

After clicking on the new entry in the list box, the spectral index is computed for every pixel for the image using default definition of A and B spectral ranges. Pixel intensities are integrated (summed) in the specified A and B spectral ranges. For example, R_{1451:1489} (in LaTeX notation) means integrated reflectance between 1451 and 1489 nm.


Each of the spectral feature parameters (in our case A and B) are also highlighted by colored bars in the spectral plot. Note that the equation is displayed in the Visualization panel using wavelength definition in nanometers. The parameters listed can be also changed directly by specifying the wavelengths.


When hovering over the image, the floating point spectral feature value is displayed in the status bar.