Symmetry based indexing of diatoms in an image database

We introduce several methods for symmetry detection and present results of their experimental evaluation on diatom images. The methods are based on distance lists, cyclic string matching, and gray level gradient direction histograms. From the experimental results it can be concluded that symmetry information is useful for various indexing tasks in image databases. Types of symmetry considered in this paper are rotational and reflectional symmetry of the shape as well as symmetry of the gray level distribution of the internal structure of objects. Using the symmetry based indexing scheme proposed, the effort for the subsequent task of diatom classification can significantly be reduced