Abstract

A semantic image description model based on generalized set is proposed, and the semantic similarity (distance) measure between images is presented. Semantic image information can be completely represented in this model as compared with previous researches based on vector space. The semantic image description model based on generalized set is similar to human understanding of image knowledge. For the purpose of the semantic image classification, semantic distance based on support vector machine classifier is employed. Experimental results show the validity of new method, and that the image classification accuracy is improved.