The visual goodness evaluation of color-based retrieval processes

A content-based image retrieval system helps users to find suitable images or videos from large databases for their purposes. In this paper we present the results of analyses of visual outputs from our image retrieval system, based on color matching. These analyses are based on human evaluation of color content similarity. Computer-based simulation results have been analyzed visually with the opinion scores method, i.e. how good the results are according to the human visual system. Tests have shown that there is no superior retrieval method that is best in every case. The goodness results of each retrieval method will be presented in this paper. The usability and usefulness of each system will be presented as well as a possible utilization domain.