Figure 5.15: Local structure and region adjacency graph.

This is a nonlinear relaxation problem. Taking eq. (5.20) into eq. (5.21), a global optimization function can be obtained:

F= k=1 T l=1 N P( q i = w k ) j=1 N l=1 T r( q i = w k , q j = w l )P( q j = w l ) ( 5.23 )

The constraints for solutions

k=1 T P( q l = w k ) =1iP( q i = w k )0i,k( 5.24 )

In the concrete use of probabilistic relaxation labeling method, the conditional probabilities of a tag for all objects are first determined, then the iteration repeats and the following ...

Get Image Understanding now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.