Basically you take a picture and you can set certain parts of the picture that do not have to match. So basically that one region of the picture can be anything as long as the rest of the picture matches. The reasoning behind this is things like radio buttons or dropdowns that could be different all the time. So instead of taking lots of little pictures around these objects you can take one larger picture.
If an Eggplant image is set to use the tolerant search type then Eggplant will ignore all pixels that have a transparency in the Alpha channel (anything not fully opaque).
The Image view in the Eggplant Suite Editor doesn’t have a facility to edit at that level so you will need to use a third party image editing tool to do this (something like Graphic Converter or Adobe Illustrator).