Semantic classification based clothing image retrieval

Performance of content-based image retrieval technology used in clothing image retrieval cannot meet user´s requirement due to the `semantic gap´ between low-level visual features and the richness of human semantics and it´s time consuming to searching images in a growing number on the internet. In order to reduce the `semantic gap´ and improve the searching speed, this paper introduces a novel clothing image retrieval method based on color semantic classification. In this method, database images are classified into different categories with color semantic features and indexed by their visual features. In the retrieval process, images with the same semantic category are selected as candidates and further ranked by their visual features. The sematic information can also be used integrated with text information labeled on the internet. Experimental results demonstrate our proposed method provides promising retrieval performance in less time than previous methods.