New content based image retrieval database structure using query by approximate shapes

The image retrieval from multimedia databases is a very challenging problem nowadays. Not only it requires the proper query form, but also efficient methods of data storage. The problem is important, because nowadays there are many different systems which needs image retrieval. As an example web searching engines may be given, which had to store a very huge amount of images and needs fast image retrieval of chosen ones. Also social media portals increasingly face the same requirements. This paper presents a new Content Based Image Retrieval database. It is based on new object representation which is based on approximation of objects by a set of shapes. The structure of the database is designed in order to reduce the number of comparisons using a tree structure. The main advantages of the proposed solution are: easy queries for users, faster image retrieval and ability to parallelize queries.