Narrowing the semantic gap in natural images

Over two decades of intensive research, researchers employed many different approaches to solve the what so called “semantic gap” problem in image indexing and retrieval. Yet, the gap is still recognized as a barrier to progress, and further work is needed to bridge (or at least) narrow the gap. This suggests that more emphasis should be placed on understanding how humans perceive images. This study measures the effectiveness of two image indexing techniques for estimating similarities between images; the semantic basis functions and the affective basis functions. These indexing techniques aim at providing a measure of similarity between outdoor natural images as humans see it. The results presented in this study suggest that the semantic basis functions outperform the affective basis functions, and are able to index the content of outdoor natural images in a manner that allows retrieval of images that have been judged to be subjectively similar.