Aggregation of color and shape features for hybrid query generation in content based visual information retrieval

A fuzzy approach for the aggregation of multiple features in content-based image retrieval is outlined. Color, shape and spatial features extracted using both computational and manual segmentation techniques are used for subsequent generation of hybrid queries to a ground truth image database consisting of architectural photographs. Retrieval results for multiple-feature queries are shown in the form of precision recall graphs. The results indicate that the fuzzy approach presented herein can perform at least as well as a weighted mean approach.