Query Decomposition: A Multiple Neighborhood Approach to Relevance Feedback Processing in Content-based Image Retrieval

Abstract

Today's Content-Based Image Retrieval (CBIR) techniques are based on the "k-nearest neighbors" (k-NN) model.They retrieve images from a single neighborhood using low-level visual features. In this model, semantically similar images are assumed to be clustered in the high-dimensional feature space. Unfortunately, no visual-based feature vector is sufficient to facilitate perfect semantic clustering; and semantically similar images with different appearances are always clustered into distinct neighborhoods in the feature space. Confinement of the search results to a single neighborhood is an inherent limitation of the k-NN techniques. In this paper we consider a new image retrieval paradigm -the Query Decomposition modelthat facilitates retrieval of semantically similar images from multiple neighborhoods in the feature space. The retrieval results are the k most similar images from different relevant clusters. We introduce a prototype, and present experimental results to illustrate the effectiveness and efficiency of this new approach to content-based image retrieval.

Một Phương Pháp Tra Cứu Ảnh Biểu Diễn Nhu Cầu Thông Tin Người Dùng Hiệu Quả

“…Nhiều hệ thống CBIR đã đƣợc phát triển, gồm QBIC [19], Photobook [4], MARS [25] NeTra [23], PicHunter [18] , Blobworld [6], VisualSEEK [28], SIMPLIcity [22] và những hệ thống khác [15,32,17,16,20,24,26,21]. Trong một hệ thống CBIR tiêu biểu, các đặc trƣng ảnh trực quan mức thấp (tức là màu, kết cấu và hình dạng) đƣợc trích rút tự động cho mục tiêu đánh chỉ số và mô tả ảnh.…”

Fast Query Point Movement Techniques for Large CBIR Systems

“…In multiple-point movement techniques such as Query Expansion [8], Qcluster [20], and Query Decomposition [17], multiple query points are used to define the ideal space that is most likely to contain relevant results. Query Expansion groups query points into clusters and chooses their centroids as Q r 's representatives (see Figure 4.1).…”

“…It uses an aggregate distance function to estimate the (dis)similarity of an object to a set of desirable images. To bridge the semantic gap more effectively, we recently proposed Query Decomposition [17]. Based on the user's relevance feedback, this scheme automatically decomposes a given query into localized subqueries, which more accurately capture images with similar semantics but in very different appearance (e.g., the front view and side view of a car), see Figure 4.3.…”

“…Many CBIR systems have been developed, including QBIC [12], Photobook [27], MARS [26], [30], NeTra [22], PicHunter [10], Blobworld [7], VisualSEEK [34], SIMPLIcity [39] and others [2], [6], [9], [14], [17], [24], [32], [38]. In a typical CBIR system, low-level visual image features (e.g., color, texture and shape) are automatically extracted for image descriptions and indexing purposes.…”

Image Retrieval Based on User-Specified Features in Multi-Cluster Queries

“…We extend this approach to handle multi-cluster queries. Recent research shows that decomposing complex queries into multiple clusters of (sub)queries yields higher precision than the traditional one-cluster methods [3].…”