FAST: Toward more effective and efficient image retrieval

Abstract

This paper focuses on developing a Fast And Semantics-Tailored (FAST) image retrieval methodology. Specifically, the contributions of FAST methodology to the CBIR literature include: (1) development of a new indexing method based on fuzzy logic to incorporate color, texture, and shape information into a region-based approach to improving the retrieval effectiveness and robustness;(2) development of a new hierarchical indexing structure and the corresponding hierarchical, elimination-based A* retrieval (HEAR) algorithm to significantly improve the retrieval efficiency without sacrificing the retrieval effectiveness; it is shown that HEAR is guaranteed to deliver a logarithm search in the average case; (3) employment of user relevance feedback to tailor the effective retrieval to each user's individualized query preference through the novel indexing tree pruning (ITP) and adaptive region weight updating (ARWU) algorithms. Theoretical analysis and experimental evaluations show that FAST methodology holds great promise in delivering fast and semantics-tailored image retrieval in CBIR.

Multimodal Retrieval with Diversification and Relevance Feedback for Tourist Attraction Images

“…To reduce the number of images to be returned to the user, some papers in the past years proposed the use of image clustering techniques [41]. ese approaches exploit the hierarchical indexing structure of the clusters to re ne the number of images to consider [28,45,50]. In [21], the authors exploit the users' feedback to modify the centroid of the considered clusters, but not to re ne the number and the shape of the clusters.…”

How to Cope with the Performance Gap in Content-Based Image Retrieval Systems

“…Although maintaining the image and its features together is interesting to transport the image, for example to the analyst's workstation, or to keep personal case records, executing content-based retrieval over a set of large images is troublesome and not efficient. Therefore, the best approach is to keep the image data set and the corresponding set of feature vectors as tables inside the storage structure of a DBMS (Barioni et al, 2006;Guliato et al, 2008;Lew et al, 2006;Zhang & Zhang, 2005). This approach guarantees the consistency of the data due to the centralized control of the DBMS, and it also allows creating the indexing structures that speed up query answering.…”

“…Liu et al, 2007). Closely related to the way that the analyst feedbacks the relevance of each image is the mathematical model of how to correct further queries to take into account the user's notion of relevance (Lim, Wu, Singh, & Narasimhalu, 2001;Rocchio, 1971;Zhang & Zhang, 2005). Another important issue is how the subsequent queries can be corrected, and the proposed techniques include modifying the similarity function (Aksoy, Haralick, Cheikh, & Gabbouj, 2000) or including bias in the query expression (Razente et al, 2008).…”

Content-based image retrieval in the topic space using SOM and LDA

“…For example, SOM and LDA have been used together for a content-based recommendation system [10]. In [14] and [15], SOM and a specific probabilistic model were combined as follows: SOM was used to compute tokens from the feature space and a pLSA-like probabilistic model was used to discover a hidden semantic information. The resulting system takes as input an image and infers a semantic representation composed of posterior probabilities on the estimated concepts.…”