An MX-CIF quadtree is a variant of quadtree which is for efficient spatial query such as whether objects are included by a spatial area [6]. when query objects are indexed, a primary result with candidates which may interect the query rectangle will be reported to have a successional precise inspection. we saved time from inspecting each of the objects exact query; the less the time is used to accomplish a query. in this paper, we propose an improve MX-CIF quadtree, compared with the original MX-CIF quadtree. a filter with our structure will decrease the failure rate of result, that is, a query will get fewer uncertain objects, the mechanism of which accelerates the secondary query. compare to original MX-CIF quadtree, with polygon data given by JTS topology suite (JTS) [2]. 42.1%-67.5% incorrectt result were filtered our by our improved MX-CIF quadtree, and its cost of tree-building time is only slightly higher than the original MX-CIF quadtree.