What is the difference between a 'Pixel-based' and 'Map-based' Arbitrary projections?

What is the difference between 'pixel based' and 'map based' arbitrary projections as defined in the Map Info of an ENVI header file? For example, when performing a Layer Stack of images that are in an arbitrary projection, it is necessary that they are in 'map based' and not a 'pixel based' type of arbitrary projection. 'Pixel based' images result in images that contain all 0's.

The "map based" arbitrary projection means that the projection is treated like any other true projection, where Y values increase as you move up/north. In a "pixel-based" arbitrary projection, Y values increase as you move down/south, in the same way the pixel rows increase down. A "pixel-based" aribitrary projection would be used when there is no true map system, but still necessary to match up with the pixel locations in your image. Images with pixel-based arbitrary projections typically will not work in Layer Stacking because Layer Stacking is for map-based projections only. However, if all of the images were in a "pixel-based" arbitrary projection and aligned in samples and lines, then it potentially could work.

Review on 12/31/2013 MM