Semantic Multi-Query Model for Cultural Computing of Image Search System

The proliferation of digital images on the internet has increased the need for image search systems, especially for culturally significant images that contain a collection of impressions. However, traditional image search systems typically rely on a single query, making it difficult to discern user intent accurately. This paper introduces a novel model for describing user impressions using a semantic multi-query function for cultural computing in image search systems. This model provides a culture-centric semantic multi-image query system to generate representative query impressions. The proposed multi-query model provides an analytical tool to semantically construct representative query color attributes, involving four stages: (1) Local normalization of 3D-Color Vector Quantization, (2) Color distribution measurement, (3) Adaptive representative color adjustment, and (4) Representative color identification. For the experimental study, we evaluate our system with two types of experiments: (1) Multi-query image for image search to ensure that our multi-query model enhances the accuracy of the retrieval outcomes, and (2) Multi-query image for semantic image search of cultural paintings. In the first experiment using the SIMLIcity dataset, our proposed multi-query model achieved better retrieval performance across most categories, reducing the single-query error from 26.67% to 20%. In the second experiment using the Indonesian cultural painting dataset, our proposed multi-query model achieved better retrieval performance across most categories, improving the single-query average similarity from 46.6% to 72%.