Object segmentation defines spatially coherent objects based on pixel classification results. In perClass Mira, many operations and analysis types extend naturally to objects. For example, we may wish to compute spectral index only on objects of interest, model quality of some objects using regression analysis, or export mean spectrum per object for further research.


The first step to apply object segmentation is to define one or more foreground classes. This can be done using Classes context menu and selecting Flag class as foreground. Alternatively, the f (for Foreground) key toggles the foreground flag for the currently selected class in the Classes list.


The Objects panel presents a separate section where objects are defined and one where objects are classified.