Skip to Main Content
In this paper, we implement a Google image search refinement that utilizes local features to find Chinese characters in search results, including following stages: (1) Chinese characters images and their SIFT features (SDB) are generated offline, (2) A text-based image search results are retrieved from the Google, (3) SIFT features of results are matched to query-text SDB using MPLSH, (4) A geometric verification algorithm is used to find the query-text and rerank results. Experiment results show that our approach is simple and effective in recognition of text in natural images, and is helpful to refine the web image search.
Computer Science and Automation Engineering (CSAE), 2012 IEEE International Conference on (Volume:1 )
Date of Conference: 25-27 May 2012