A Semantic Multi-Query Image Search System with Analytical Function for Representative Query Color Generation

The effective analyzing query method is indispensable to image search system that is needed to retrieve appropriate images. The image search system commonly uses a content based search for the retrieval with a single query. In case of culture related images those treasure set of impressions behind the feature contents, the image search needs to provide more flexible query model to apprehend what the user wants to retrieve. This paper presents a new model for representing the user impressions by providing a semantic multi-query image search system. The proposed multi-query model provides an analytical function to semantically generate the representative query color features. The function consists of four steps: (1) The normalized 3D-Color Vector Quantization for local extraction of color features in an image, (2) Measurement of color distribution among image queries by calculating average and standard deviation of extracted color features, (3) Adaptive adjustment of representative colors by measuring the density of the color distribution, and (4) Identification of representative colors by applying cluster based similarity measurement of the color density. To perform our proposed semantic multi-query image search system, we examine our system with a SIMLIcity Dataset containing 1000 images with 10 categories. We also apply our semantic multi-query model to the Indonesian cultural paintings for the applicability of the culture based image search. Keywords: image search system, semantic multi-query model, representative color identification.