This paper proposed that the design of fast search path for image matching should be done right in the feature space of the reference image. By rearranging the feature space into an orderly structure

we can have a global grasp of its distribution.In this way

we are no longer restricted to the local information that traditional fast search algorithms can only rely on

hence the problem of local minima is avoided.In the orderly feature space

binary search can be adopted as the core of a new fast search algorithm.Results from simulated experiments have shown that the new algorithm can yield excellent searching speed and very good matching accuracy.