Detecting Regular Structures for Invariant Retrieval

Many of the existing approaches to invariant content-based image retrieval rely on local features, such as color or specific intensity patterns (interest points). In some methods, structural content is introduced by using particular spatial configurations of these features, which are typical for the pattern considered. Such approaches are limited in their capability to deal with regular structures when high degree of invariance is required. Recently, we have proposed a general measure of pattern regularity [2] that is stable under weak perspective of non-flat patterns and varying illumination. In this paper we apply this measure to invariant detection of regular structures in aerial imagery.