A Semi-Automatic Object Extraction Tool for Querying in Multimedia Databases Ediz Saykol, Ugur Güdükbay, and Özgür Ulusoy   View Paper (PDF)   Return to Papers Abstract Considering the complexity and huge volume of image and/or video data, efficient methods need to be developed forquerying multimedia databases. A well-known technique used for querying multimedia data is query-by-feature (e.g. color, shape, texture, size) of the objects residing in images and/or video frames. In this paper, we propose a tool for extracting objects from images and/or video frames, called Object Extractor, as well as the ways of coping with the object features within extracted objects. The tool is semi-automatic in the sense that the user specifies the colors on the object by clicking the mouse to make the tool capture object pixels. In order to extract objects, an improved version of the Flood Fill algorithm for polygon filling is provided. The extraction algorithm uses filtered images to perform better. Moreover, the experimental results obtained for evaluating the performance of the tool in extracting objects are presented. It is shown through these results that a few mouse clicks would suffice to extract objects effectively. DiSC'02 © 2003 Association for Computing Machinery