Forsyth, D. A., Malik, J., and Wilensky, R. Searching for Digital Pictures. Scientific American, June 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....IP 4.2 (Image Search and Sorting) I. INTRODUCTION Searching for digital information, especially images, music, and video, is quickly gaining in importance for business and entertainment. Content based image retrieval (CBIR) is receiving widespread research interest [1] 4] 2] 3] 7] [8], 9] 10] 11] 12] 13] 14] 6] 15] 16] 17] 18] 19] 20] It is motivated by the fast growth of image databases which, in turn, require efficient search schemes. A search typically consists of a query followed by repeated relevance feedback, where the user comments on the items ....

D. Forsyth, J. Malik, and R. Wilensky, "Searching for digital pictures," Scientific American, pp. 88-93, 1997.

....images of horses, for example, in a variety of poses. Fleck et al. [8] use knowledge about the positions of attachment of limbs and head to the human body to detect the presence of naked people in the database images. Forsyth et al. illustrate some specialized applications of image retrieval in [9]. There has been a lot of work in the area of image segmentation. Recent work has focused on combination of different cues like color, texture and edges for segmentation [2, 20, 19] Re lational graph matching has been used for segmenting natural images in [31] However, these techniques produce ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, 276(6):72- 77, Jun 1997.

....measure, visual parts, discrete curve evolution. i Introduction With the recent increase in image and multimedia databases, there has been an acceler ation of research in developing and applying shape similarity measures, e.g. for shape based retrieval of similar objects (see Forsyth et al. [1]) In computer vision there is a long history of work on shape representation and shape similarity. However, most of the existing methods have only a very limited possible application to distributed image databases, since the shape of objects must be restricted and known a priori. These meth ods ....

....large number of vertices without loss of information. We assign to a every polygonal curve a tangent function, which is a step function. We use the tangent function as a basis for the proposed similarity measure of simple polygonal arcs. Let C be a polygonal curve. We treat it as a function C: [0, 1] IR 2, i.e. the length of C is rescaled to 1. The tangent function of C (which is also called a turning function) is a multi valued function T(C) 0,1] 0, 2r] defined by T(C) s) C (s) and T(C) s) C (s) where C (s) and C (s) are left and right derivatives of C. For example, see Figures ....

[Article contains additional citation context not shown here]

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, pages 88-93, June 1997.

....their area peak basically removing noise that is small amounts of isolated colours. Color WISE uses Microsoft Access to perform the database functions, and uses a similarity metric based on IBM s QBIC system. Querying in Color WISE is achieved with query by image. The Digital Library Project [5] taking place at the University of Berkeley, California, uses low level grouping techniques to create blobs of stuff , which can be texture, colour, or symmetry. The blobs can be matched against their content, and their position, and it is possible to use high level techniques to analyse the ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, pages 72--77, July 1997.

....data are very large; high speed networking support is required to allow interaction with and collaboration using such objects at reasonable rates. Retrieving images from very large collections using image content as a key is an important component of our research, described in more detail in [36, 37]. Users are primarily interested in scenes containing particular objects or configurations of objects. We have already defined a representation of images as collections of blobs of coherent color and texture the blob world representation [12] Blob world is easier to use than other current image ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, July 1997.

....IP 4.2 (Image Search and Sorting) I. Introduction Searching for digital information, especially images, music, and video, is quickly gaining in importance for business and entertainment. Content based image retrieval (CBIR) is receiving widespread research interest [1] 4] 2] 3] 7] [8], 9] 10] 11] 12] 13] 14] 6] 15] 16] 17] 18] 19] 20] It is motivated by the fast growth of image databases which, in turn, require ecient search schemes. A search typically consists of a query followed by repeated relevance feedback, where the user comments on the items ....

D. Forsyth, J. Malik, and R. Wilensky, \Searching for digital pictures," Scientic American, pp. 88-93, 1997.

....a way analogous to keyword searches of text databases. The most natural and practically achievable way for users to initiate such a search is via a query based on image content like find images which look like this one or find images with this texture or other more complex and compound queries [7]. Consequently, automatic image indexing and retrieval based on image content has become an important and active area of research. In this paper we review this relatively new area of research from a high level (in this section) and also provide a more detailed look at some specific image retrieval ....

....possible level of abstraction computers cannot currently understand images at the semantic level. So an image retrieval system based on semantics is not possible 3 [6] To a limited extent, image retrieval based on the object level is 2 The ideas in this subsection are drawn primarily from [7, 29, 37]. 3 This does not prevent authors from using the word semantic in describing their image retrieval approach. What they typically mean is object (or geometric ) which is one level down in the hierarchy. currently possible. This usually is done by restricting attention to images with just a ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, pages 88--93, June 1997.

....to decompose areas of images into partitions with coherent colour and texture properties. Collections of spatially associated regions of similar properties are used to compose objects in the scene and the data is employed to identify similarly orientated collections of blobs in other scenes [4]. One of the most significant contributors is MIT s Media Laboratory whose FourEyes system, which builds on the work surrounding the well known PhotoBook project [13] tries to overcome the difficulties of dimensional explosion in feature space by using a society of models [14] In an initial ....

....algorithm [10] This process results in irregularly shaped regions which should represent the actual outlines of objects in the images. It is on these areas that all queries are based as opposed to either the entire image or rectangular minimum bounding boxes as employed in some approaches [11][4][14] This is an important distinction since the feature extraction will not draw any contaminating biases from outside the required object. With a segmentation available, a feature extraction process is used to obtain information unique to each region. The components extracted are as follows: 0 ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for Digital Pictures. Scientific American, pages 72--77, June 1997.

....has been on how accurate the match is, or in other words how similar the retrieved image is to the original image. For example a query asking for an image of the sun should not be answered with an image containing an orange ball. The focus has been to extract more descriptive features [WK96, FMW97, LT96, KCH95] to represent extracted features more accurately [NS96, Lah95] and to represent more relationships among image objects [GWJ91, GOC 92] Each of these objectives has been studied with the goal of matching two images accurately. 2.4 Prototype Systems We describe prototype systems ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, pages 88--93, June 1997.

No context found.

Forsyth, D. A., Malik, J., and Wilensky, R. Searching for Digital Pictures. Scientific American, June 1997.

....kept pace with the collections they are searching. The shortcomings of these systems are due both to the image representations they use and to their methods of accessing those representations to find images: While users generally want to find images containing particular objects ( things ) 4, 6] most existing image retrieval systems represent images based only on their low level features ( stuff ) with little regard for the spatial organization of those features. 675 Systems based on user querying are often unintuitive and offer little help in understanding why certain images were ....

....representations to find images: While users generally want to find images containing particular objects ( things ) 4, 6] most existing image retrieval systems represent images based only on their low level features ( stuff ) with little regard for the spatial organization of those features. 675 Systems based on user querying are often unintuitive and offer little help in understanding why certain images were returned and how to refine the query. Often the user knows only that he has submitted a query for, say, a bear and retrieved very few pictures of bears in return. For general ....

[Article contains additional citation context not shown here]

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, 276(6):72 77, June 1997.

.... the understandability of results and ease of refining the query. All of these factors should be considered together when designing a system. In addition, image database users generally want to find images based on the objects they contain, not just low level features such as color and texture [5, 7]; image retrieval systems should be evaluated based on their performance at this task. Current image retrieval systems tend to perform queries quickly but do not succeed in the other two areas. A key reason the quality of query results suffers is because the systems do not look for meaningful ....

Forsyth, D., Malik, J., Wilensky, R.: Searching for digital pictures. Scientific American, 276 (June 1997) 72--77

.... the understandability of results and ease of refining the query. All of these factors should be considered together when designing a system. In addition, image database users generally want to find images based on the objects they contain, not just low level features such as color and texture [5, 7]; image retrieval systems should be evaluated based on their performance at this task. Current image retrieval systems tend to perform queries quickly but do not succeed in the other two areas. A key reason for the poor quality of query results is that the systems do not look for meaningful image ....

Forsyth, D., Malik, J., Wilensky, R.: Searching for digital pictures. Scientific American, 276 (June 1997) 72--77

....have not kept pace with the collections they are searching. The limitations of these systems include both the image representations they use and their methods of accessing those representations to find images: ffl While users generally want to find images containing particular objects ( things ) [9, 13], most existing image retrieval systems represent images based only on This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version will be superseded. their low level features ( stuff ) with little regard for the ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, 276(6):72--77, June 1997.

....kept pace with the collections they are searching. The shortcomings of these systems are due both to the image representations they use and to their methods of accessing those representations to find images: ffl While users generally want to find images containing particular objects ( things ) [4, 6], most existing image retrieval systems represent images based only on Proc. Int. Conf. Comp. Vision 1998. Copyright (c) 1998 IEEE. their low level features ( stuff ) with little regard for the spatial organization of those features. ffl Systems based on user querying are often unintuitive ....

D. Forsyth, J. Malik, and R. Wilensky. Searching for digital pictures. Scientific American, 276(6):72--77, June 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC