A new SVM-based active feedback scheme for image retrieval. (January 2015)

• Record Type: Journal Article
• Authors: Wang, Xiang-Yang; Yang, Hong-Ying; Li, Yong-Wei; Li, Wei-Yi; Chen, Jing-Wei
• Is Part Of: Engineering applications of artificial intelligence. Volume 37(2015:Jan.)
• Page Start: 43
• Page End: 53
• Publication Date: 2015-01
• Subjects: Content-based image retrieval -- Relevance feedback -- Support vector machine -- Active learning -- Ensemble classifiers -- Output codes; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2014.08.012
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Relevance feedback has emerged as a powerful tool to boost the retrieval performance in content-based image retrieval (CBIR). Support vector machine (SVM) active learning is one popular and successful technique for relevance feedback in CBIR. Despite the success, for conventional SVM active learning, the users are usually not so patience to label a large number of training instances in the relevance feedback round. To overcome this limitation, a new SVM-based active feedback using ensemble multiple classifiers is proposed in this paper. Firstly, we select the most informative images by using active learning method for user to label, and quickly learn a boundary that separates the images that satisfy the user׳s query concept from the rest of the dataset. Then, a set of moderate accurate one-class SVM classifiers are trained separately by using different sub-features vectors. Finally, we compute the weight vector of component SVM classifiers dynamically by using the parameters for positive and negative samples, and combine the results of the component classifiers to form an output code as a hypothesized solution to the overall image retrieval problem. Experiments on large databases show that the proposed algorithms are significantly more effective than the state-of-the-art approaches.