Semantic annotation method of clothing image

Abstract

Semantic annotation is an essential issue for image retrieval. In this paper, we take the online clothing product images as sample. In order to annotate images. we first segment the image into regions, then remove the background and noise information. The illumination and light interference is considered too. Cloth position and region are determined by rules. Images are translated into some features. Visual words are prepared by human and calculate methods. Finally, Image features are mapped to different visual words. Pre-processing and post-processing steps which uses face recognition method and background rule analysis are applied. Finally, some segmentation and annotation results are given to discuss the method. © 2013 Springer-Verlag.