When all paths to all candidate points have been obtained, we evaluate the quality of each path which depends on several criterions. For each candidate point, a value is allocated, that is a weighted mean of criterions values. All these values belong to [0,1] interval, and the highest gives the best candidate point.

The criterions are the followings:

• First criterion is the radiometry homogeneity of the path. Standard deviation of radiometry is computed:

• The value 1 corresponds to the smallest deviation and
• 0 to the highest one.
  
• Second criterion is the cost of the path returned by F* algorithm. The value returned by this criterion is defined identically to the previous one.
  

The general value of the path is the mean of the two values returned by the two criterions:

• If it is 0 then tracking stops,
• else, it stops when the starting point is near by the image edge.

Search window must be large enough for a good result of F* algorithm but not too much, because the searched linear feature is considered as crossing the last column of search window. This method gives good results, is fast and needs little memory (for search window mainly) whatever the size of the image.