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.

Image feature space is typically complex due to the high dimensionality of data. Effective handling of this space has prompted many research efforts in the study of dimensionality reduction in the image domain. In this paper, we propose a semisupervised reduction method that leverages relevance feedback information in the retrieval process to learn suitable linear and orthogonal embeddings. In the reduced space constructed by the proposed embedding, relevant images are kept close to each other, while irrelevant ones are dispersed far apart. The experimental results demonstrate the superiority of our method.

Target search in content-based image retrieval (CBIR) systems refers to finding a specific (target) image such as a particular registered logo or a specific historical photograph. Existing techniques, designed around query refinement based on relevance feedback, suffer from slow convergence, and do not guarantee to find intended targets. To address these limitations, we propose several efficient query point movement methods. We prove that our approach is able to reach any given target image with fewer iterations in the worst and average cases. We propose a new index structure and query processing technique to improve retrieval effectiveness and efficiency. We also consider strategies to minimize the effects of users' inaccurate relevance feedback. Extensive experiments in simulated and realistic environments show that our approach significantly reduces the number of required iterations and improves overall retrieval performance. The experimental results also confirm that our approach can always retrieve intended targets even with poor selection of initial query points.

Mobile ad-hoc networks (MANETs) have attracted great research interest in recent years. Among many issues, lack of motivation for participating nodes to collaborate forms a major obstacle to the adoption of MANETs. Many contemporary collaboration enforcement techniques employ reputation mechanisms for nodes to avoid and penalize malicious participants. Reputation information is propagated among participants and updated based on complicated trust relationships to thwart false accusation of benign nodes. The aforementioned strategy suffers from low scalability and is likely to be exploited by adversaries. In this paper, we propose a novel approach to address these problems. With the proposed technique, no reputation information is propagated in the network and malicious nodes cannot cause false penalty to benign hosts. Nodes classify their one-hop neighbors through direct observation and misbehaving nodes are penalized within their localities. Data packets are dynamically rerouted to circumvent selfish nodes. As a result, overall network performance is greatly enhanced. This approach significantly simplifies the collaboration enforcement process, incurs low overhead, and is robust against various malicious behaviors. Simulation results based on different system configurations indicate that the proposed technique can significantly improve network performance with very low communication cost.

Recent content-based image retrieval (CBIR) techniques were designed around query refinement based on relevance feedback. They suffer from slow convergence, high disk I/O, and do not even guarantee to find intended targets. In this paper, we identify the cause of these problems and propose several efficient target search methods to address these drawbacks. Our complexity analysis shows that our approach is able to reach any given target image with fewer iterations in the worst and average cases. We evaluated our techniques on large datasets in simulated and realistic environments. The results show that our approach significantly reduces the number of iterations and improves overall retrieval performance. The experiments also confirm that our approach can always retrieve intended targets even with poor selection of initial query points and can be used to improve the effectiveness of existing CBIR systems with relevance feedback.