An Enhanced Bag-of-Visual Word Vector Space Model to Represent Visual Content in Athletics Images

摘要: Images that have a different visual appearance may be semantically related using higher level conceptualization. However, image classification and retrieval systems tend to rely only on the low-level structure within images. This paper presents framework deal with this semantic gap limitation by exploiting well-known bag-of-visual words (BVW) represent content. The novelty of is threefold. First, quality improved constructing from representative keypoints. Second, domain specific “non-informative words” are detected which useless content data but can degrade categorization capability. Distinct existing frameworks, two main characteristics for non-informative defined: high document frequency (DF) small statistical association all concepts in collection. third contribution novel method used restructure vector space model respect structural ontology order resolve synonym polysemy problems. experimental results show our disambiguate word senses effectively significantly improve classification, interpretation, performance athletics