IMPROVING CONTENT-BASED RETRIEVAL ON ELECTRONIC APPAREL CATALOG RETRIEVAL SYSTEM

摘要

Content-based image retrieval (CBIR) methods are useful tools online e-commerce applications, such as in electronic apparel catalog retrieval system (EACRS). In this paper, we investigate a CBIR based electronic hat catalog retrieval system and conduct an empirical study for a specific e-commerce application. Through exploring three types of low-level features: texture, color and shape, we identify appropriate low-level features for semantic retrieval of this specific type of hat catalog database. The experimental results indicate that the shape features based on the centroid-contour distance Fourier descriptor perform better than the color and texture features for the hat database used in this project.