Interpreting visual meaning: a comparative study of concept-based and content-based image indexing

Concept-based and content-based image indexing were evaluated using text search and image search on search engines such as Google Images and Bing Visual Search. The findings showed that the effectiveness of each indexing method depends on the user’s search intent and the type of image being sought – abstract concepts, objects, scenery and locations. A trade-off emerged between the specificity and the exhaustivity of the retrieved results, along with key factors that could be refined to improve search accuracy. To address these limitations, a hybrid model is proposed that combines the expansive, idea-generating strengths of one approach with the focused precision of the other. Search engines with AI image analysis can largely assist but not replace human-led subject analysis and indexing.