A multimedia database server for information storage and querying

The paper presents an original software tool for creating and querying medium sized image collections. The software tool includes a server for a relational multi-user MMDBMS that uses traditional data types (integer, double, char, varchar) and a complex data type, called image. This type is used for storing the image and the extracted information regarding the size and type of the image, color and texture characteristics. An element of originality for this software is the content based retrieval module that allows the user to build content based visual queries to the image level. The color information is represented by the color histograms automatically extracted from the image. The software uses HSV color space, quantized to 166 colors. The dissimilitude between images is calculated using one of the following metrics: Euclidian metrics, histogram intersection or a combination of them. One of the advantages of this tool is that it has a low cost and it is easy to maintain, comparing to other similar database servers.