The application fields, such as CAD and GIS, that require spatial data processing are expanding rapidly and it is important to manage the spatial data efficiently using database systems(DBS). The spatial DBS must support spatial queries and join operations efficiently. Due to the complexity of processing spatial joins, the spatial join operation is divided into filter and refinement steps in general. It is to filter the objects that would not join together using the approximation of objects in advance and then to refine candidates using expensive computational geometry algorithms. We select MBR-join as a filter step method, and TRAP and DMBR object decomposition approaches as refinement step methods. And then, we verify the feasibility of implementation under a real spatial DBS environment and measure performance of the implemented filter and refinement methods. In addition, we compare the performance results and object decomposition overheads between the two refinement methods.
