Efficient evaluation of spatial queries is an important issue in spatial database. Among spatial operations

spatial join is veryuseful

intersection being the most common predicate. However

the exact intersection test of two spatial objects is the most timeconsuming and I/O-consuming step in processing spatial joins. The use of approximations can reduce the need for examining the exact geometry of spatial objects in order to findthe intersecting ones. This work proposes a new raster approximation （Three-Color Raster Signature -3CRS） for representing dif- ferent data types （polygons

polylines and points）

and to be used as filter the Multi-Step Query Processor. We have also executed experimental tests over realdatasets

the results having demonstrated the effectiveness of our approach.