Aya Soffer

MARCO: MAp retrieval by COntent

A system named MARCO (denoting map retrieval by content) that is used for the acquisition, storage, indexing, and retrieval of map images is presented. The input to MARCO are raster images of separate map layers and raster images of map composites. A legend-driven map interpretation system converts map layer images from their physical representation to their logical representation. This logical representation is then used to automatically index both the composite and the layer images. Methods for incorporating logical and physical layer images as well as composite images into the framework of a relational database management system are described. Indices are constructed on both the contextual and the spatial data thereby enabling efficient retrieval of layer and composite images based on contextual as well as spatial specifications. Example queries and query processing strategies using these indices are described. The user interface is demonstrated via the execution of an example query. Results of an experimental study on a large amount of data are presented. The system is evaluated in terms of accuracy and in terms of query execution time.

Integration of local and global shape analysis for logo classification

Abstract A comparison is made of global and local methods for the shape analysis of logos in an image database. The qualities of the methods are judged by using the shape signatures to define a similarity metric on the logos. As representatives for the two classes of methods, we use the negative shape method which is based on local shape information and a wavelet-based method which makes use of global information. We apply both methods to images with different kinds of degradations and examine how a particular degradation highlights the strengths and shortcomings of each method. Finally, we use these results to develop a new adaptive weighting scheme which is based on the relative performances of the two methods. This scheme gives rise to a new method that is much more robust with respect to all degradations examined and works by automatically predicting if the negative shape or wavelet method is performing better.

Using negative shape features for logo similarity matching

A method for representing and matching logos based on positive and negative shape features is presented. Negative shape features represent an object that consists of several components enclosed in a simple geometric structure (e.g., a square) based on its interior with the components considered as holes. The goal is to find logos in a database that are most similar to a given sample logo. A border is added around logos that are not enclosed in a simple shape. Logos are segmented. Local and global shape features are computed for each component. Two methods for comparing logos represented by positive and negative components are presented and evaluated.

Knowledge Agents on the Web

This paper introduces and evaluates a new paradigm, called Knowledge Agents, that incorporates agent technology into the process of domainspecific Web search. An agent is situated between the user and a search engine. It specializes in a specific domain by extracting characteristic information from search results. Domains are thus user-defined and can be of any granularity and specialty. This information is saved in a knowledge base and used in future searches. Queries are refined by the agent based on its domain-specific knowledge and the refined queries are sent to general purpose search engines. The search results are ranked based on the agent’s domain specific knowledge, thus filtering out pages which match the query but are irrelevant to the domain. A topological search of the Web for additional relevant sites is conducted from a domain-specific perspective. The combination of a broad search of the entire Web with domain-specific textual and topological scoring of results, enables the knowledge agent to find the most relevant documents for a given query within a domain of interest. The knowledge acquired by the agent is continuously updated and persistently stored thus users can benefit from search results of others in common domains.

A legend-driven geographic symbol recognition system

A system is presented that utilizes the symbolic knowledge found in the legend of the map to drive geographic symbol recognition. The geographic symbol layer(s) of the map are first scanned. The legend of the map is located and segmented. The geographic symbols are identified and their semantic meaning is attached to them. The geographical symbols in input maps are classified using statistical pattern recognition. An experimental study was conducted on a large amount of data with recognition rates of over 95%.

MAGELLAN: Map Acquisition of GEographic Labels by Legend ANalysis

Abstract. A system named MAGELLAN (denoting Map Acquisition of GEographic Labels by Legend ANalysis) is described that utilizes the symbolic knowledge found in the legend of the map to drive geographic symbol (or label) recognition. MAGELLAN first scans the geographic symbol layer(s) of the map. The legend of the map is located and segmented. The geographic symbols (i.e., labels) are identified, and their semantic meaning is attached. An initial training set library is constructed based on this information. The training set library is subsequently used to classify geographic symbols in input maps using statistical pattern recognition. User interaction is required at first to assist in constructing the training set library to account for variability in the symbols. The training set library is built dynamically by entering only instances that add information to it. MAGELLAN then proceeds to identify the geographic symbols in the input maps automatically. MAGELLAN can be fine-tuned by the user to suit specific needs. Recognition rates of over 93% were achieved in an experimental study on a large amount of data.

Image categorization using texture features

A method for finding all images from the same category as a given query image (termed categorization) using texture features is presented. The hypothesis that two images that are similar in texture are likely to belong to the same category is examined. A new texture feature called an N/spl times/M-gram is presented. It is based on the N-gram technique that is commonly used for text similarity. The process of computing an image profile in terms of its N/spl times/M-grams is described. Results of experiments on images from various categories are presented. The N/spl times/M-gram method with three different similarity measures is compared to the results of categorization using other well-known texture features and grey-level distribution features. The results show that, for our test images, texture features are suitable for image categorization, and N/spl times/M-gram based methods are the best overall choice of texture features for this task.

Pictorial queries by image similarity

A method for specifying pictorial queries to an image database is introduced. A pictorial query specification consists of a query image, and a similarity level that specifies the required extent of similarity between the query image and database images that are to be retrieved. The query image is constructed by positioning objects so that the desired locational and spatial constraints hold. Two image similarity factors are considered: (1) contextual similarity: how well does the content of one image match that of another; (2) spatial similarity: the relative locations of the matching symbols in the two images. Algorithms for retrieving all database images that conform to a given pictorial query specification are presented and compared.

Integrating symbolic images into a multimedia database system using classification and abstraction approaches

Abstract. Symbolic images are composed of a finite set of symbols that have a semantic meaning. Examples of symbolic images include maps (where the semantic meaning of the symbols is given in the legend), engineering drawings, and floor plans. Two approaches for supporting queries on symbolic-image databases that are based on image content are studied. The classification approach preprocesses all symbolic images and attaches a semantic classification and an associated certainty factor to each object that it finds in the image. The abstraction approach describes each object in the symbolic image by using a vector consisting of the values of some of its features (e.g., shape, genus, etc.). The approaches differ in the way in which responses to queries are computed. In the classification approach, images are retrieved on the basis of whether or not they contain objects that have the same classification as the objects in the query. On the other hand, in the abstraction approach, retrieval is on the basis of similarity of feature vector values of these objects. Methods of integrating these two approaches into a relational multimedia database management system so that symbolic images can be stored and retrieved based on their content are described. Schema definitions and indices that support query specifications involving spatial as well as contextual constraints are presented. Spatial constraints may be based on both locational information (e.g., distance) and relational information (e.g., north of). Different strategies for image retrieval for a number of typical queries using these approaches are described. Estimated costs are derived for these strategies. Results are reported of a comparative study of the two approaches in terms of image insertion time, storage space, retrieval accuracy, and retrieval time.

Pictorial Query Specification for Browsing Through Spatially Referenced Image Databases

Abstract A pictorial query specification technique that enables the formulation of complex pictorial queries for browsing through a collection of spatially referenced images is presented. It is distinguished from most other methods by the fact that in these methods the query image specifies a target database image in its entirety whereas in our approach the query image specifies the combination of objects that the target database image should contain rather than being treated as a whole image. The query objects are represented by shape features although other features such as color, texture or wavelets could also be used. Using our technique, it is possible to specify which particular objects should appear in the target images as well as how many occurrences of each object are required. Moreover, it is possible to specify the minimum required certainty of matching between query-image objects and database-image objects, as well as to impose spatial constraints that specify bounds on the distance between objects and the relative direction between them. These spatial constraints can also be used to specify other topological relations such as enclosure, intersection, overlap, etc. Each pictorial query is composed of one or more query images. Each query image is constructed by selecting the required query objects and positioning them according to the desired spatial configuration. Boolean combinations of two or more query images are also possible by use of AND and OR operators. A query image may be negated in order to specify conditions that should not be satisfied by the database images that are retrieved successfully. In addition, a capability is provided to specify whether the same instance of an object is to be used when it appears in more than one of the query images that make up the pictorial query, or whether two different instances are allowed. Several example queries are given that demonstrate the expressive power of this query specification method. An algorithm for retrieving all database images that conform to a given pictorial query specification is presented. The user interface for using this pictorial query specification method to browse the results in a map image database application is described and illustrated via screen shots.