Filter and Refine Strategy

Synonyms

Filter-Refine Paradigm

Definition

The filter and refine strategy is a general method for completing a computationally intensive task as quickly as possible. In some tasks, such as a spatial join, we need to do a computationally intensive step on only a small subset of the original data. In order to find this subset, we first apply a filter step which removes data that are not involved in our task. We then do the refining step on the remaining data.

Main Text

When we are using the filter and refine strategy with spatial data, we generally filter the data by running the algorithm in question on simplified representations of the spatial objects. Minimum bounding rectangles (MBRs) are commonly used for this purpose. The point of the filter step is not to reduce the data to only that data needed to find the answer to the problem, but rather to remove a significant part of the uninvolved data in a computationally efficient manner. For example, when trying to detect overlap present...