Traditional image retrieval technology is pixel sensitive and with low fault tolerance. To overcome this deficiency, a novel method for large-scale image retrieval is proposed in this paper, which is especially suitable for images with kinds of interferences, such as rotation, pixel lost, watermarks, etc. First, local features of images are extracted to build a visual dictionary with weight, which is a new data structure developed from bag-of-words. In the retrieval process, we look up all the features extracted from the target image in the dictionary and create a single list of weight to get the result. We demonstrate the effectiveness of our approach using a coral image set and online image set on eBay. © 2012 Springer-Verlag.
CITATION STYLE
Yin, C. Q., Mao, W., & Jiang, W. (2012). Very large-scale image retrieval based on local features. In Communications in Computer and Information Science (Vol. 304 CCIS, pp. 242–250). https://doi.org/10.1007/978-3-642-31837-5_36
Mendeley helps you to discover research relevant for your work.