|
0 is a place-holder value to indicate that IPP was unable to calculate the fractal dimension of the object. For closed curves, no fractal dimension should be less than 1. A zero-value will be returned when there are not enough points along the outline of the object to calculate the FD reliably using the method outlined below. If I'm not mistaken, I think the number of points must be greater than 20 or 30. A quick and dirty fix is for you to resize your images up to get bigger objects prior to the counting/sizing. A more involved fix would be to write a macro implementing John Russ' method using the Euclidean distance map (See his book). From a previous posting: ...IPP implements a variation on the hand and divider method (first introduced by Richardson). IPP accounts for variations in the perimeter measurement as a result of different starting points along the outline (i.e. depending on the where you first put the ruler to start measuring the outline, the perimeter is going to vary slightly even if the ruler stays the same). The original method doesn't compensate, which is one of its main drawbacks... Jean-Paul Martin
|