In this paper, we present the results of our work that seeks to negotiate the gap between low-level features and high-level concepts in the domain of web document retrieval. This work concerns a technique, Latent Semantic Indexing (LSI), which has been used for textual information retrieval for many years. In this environment, LSI is used to determine clusters of cooccurring keywords, sometimes, called concepts, so that a query which uses a particular keyword can then retrieve documents perhaps not containing this keyword, but containing other keywords from the same cluster. In this paper, we examine the use of this technique for content-based web document retrieval, using both keywords and image features to represent the documents. Two different approaches to image feature representation, namely, color histograms and color anglograms, are adopted and evaluated. Experimental results show that LSI, together with both textual and visual features, is able to extract the underlying semantic structure of web documents, thus helping to improve the retrieval performance significantly.

Besides traditional applications (e.g., CAD/CAM and Trademark registry), new multimedia applications such as structured video, animation, and MPEG-7 standard require the storage and management of well-defined objects. This study compared four shape-based object retrieval techniques (FD, GB, DT , and TPV AS). The similarity retrieval accuracy of our method (TV PAS) was comparable to the other methods, while it had the lowest computation cost to generate the shape signatures of the objects. Moreover, it has low storage requirement, and a comparable computation cost to compute the similarity between two shape signatures. In addition, TPV AS requires no normalization of the objects, and is the only method that has direct support to RST query types. We also introduced a new shape description taxonomy. 1 Introduction Many applications in the areas of CAD/CAM and computer graphics require to store and access large databases [32, 31]. These object databases are then queried and search...

Abstract not found

This paper examines the use of a computational geometry-based spatial color indexing methodology for efficient and effective image retrieval. In this scheme, an image is evenly divided into a number of M*N nonoverlapping blocks, and each individual block is abstracted as a unique feature point labeled with its spatial location, dominant hue, and dominant saturation. For each set of feature points labeled with the same hue or saturation, we construct a Delaunay triangulation and then compute the feature point histogram by discretizing and counting the angles produced by this triangulation. The concatenation of all these feature point histograms serves as the image index. An important contribution of this work is to encode the spatial color information using geometric triangulation, which is translation, rotation, and scale independent. We have implemented the proposed approach and have tested it over two image collections of 2000 JPEG images and 1380 GIF images. Various experimental res...

Content-based image retrieval (CBIR) has been an active research area in the last ten years, and a variety of techniques have been developed. However, retrieving images on the basis of low-level features has proven unsatisfactory, and new techniques are needed to support high-level queries. Research efforts are needed to bridge the gap between high-level semantics and low-level features. In this paper, we present a novel approach to supporting semantics-based image retrieval. Our approach is based on the monotonic tree, a derivation of the contour tree for use with discrete data. The structural elements of an image are modeled as branches (or subtrees) of the monotonic tree. These structural elements are classified and clustered on the basis of such properties as color, spatial location, harshness and shape. Each cluster corresponds to some semantic feature. This scheme is applied to the analysis and retrieval of scenery images. Comparisons of experimental results of this approach with conventional techniques using low-level features, demonstrate the effectiveness of our approach. Keywords: Content-based image retrieval, image feature extraction, annotation, semantics retrieval, monotonic tree 1

Abstract — We present a robust method for detecting and recognizing topological features in underground mines. Our method involves performing Delaunay triangulations on range scans to extract points of interest, such as intersecting corridors. By combining these interest points into a topological map, we have a valuable tool for navigation and localization in large scale, highly cyclic environments. We present results from a research coal mine near Pittsburgh, PA. I.

Shape of an object is an important feature for image and multimedia similarity retrievals. However, as a consequence of uncertainty, shape representation techniques may sometimes work well only in certain environments, and their performance may depend crucially on the quality of the technique used to represent shapes. Therefore, to address the issues of robustness and stability of the alternative shape representation techniques, it is vital to evaluate the behavior of such techniques in the presence of some uncertainty/discrepancies. In this study we focus on shape-based object retrieval under various uncertainty scenarios and conduct a comparison study on four techniques (i.e., Fourier descriptors (FD), grid based (GB), Delaunay triangulation (DT ), and one of our proposed MBC-based methods (i.e., MBC-TPV AS)). We measure the eectiveness of the similarity retrieval of the four dierent shape representation methods (in terms of recall and precision) under the following situat...

One of the main goals in Content Based Image Retrieval (CBIR) is to incorporate shape into the process in a reliable manner. In order to overcome the difficulties of directly obtaining shape information (in particular avoiding region segmentation) we develop several shape measures that tackle the problem in an indirect manner, requiring only a minimal amount of segmentation. A histogram-based scheme is then used, maintaining low complexity with high efficiency and robustness. The obtained results showed that the synergy of the shape measures worked out providing an improvement over the colour histogram.

Introduction With the advance of multimedia technology, image data in various formats are becoming available at an explosive rate. With such enormous data resources, the search and retrieval of image databases are demanded to provide open access to relevant information and products. Thus, content-based image retrieval (CBIR) has become an active research area. A variety of techniques have been developed. In particular, content-based image retrieval using lowlevel features such as color [41, 36, 26], texture [21, 35, 34, 40, 20], shape [42, 22, 14, 23, 24, 15, 10, 17, 46, 47, 33, 43, 32] and others [30, 38, 2, 16, 7] extracted from the images has been well studied. Various image querying systems including QBIC [11], VisualSeek [36], PhotoBook [27] and Virage [5] have been built based on the low-level features for general or specific image retrieval tasks. However, retrieving images based on low-level features has been proven unsatisfactory. Thus, effective and precise image retrieval

Abstract not found