#### DMCA

## Emergent Semantics Through Interaction in Image Databases (2001)

Venue: | IEEE Transactions on Knowledge and Data Engineering |

Citations: | 93 - 10 self |

### Citations

2944 |
Numerical recipes in C : the art of scientific computing
- Press, Teukolsky, et al.
- 1992
(Show Context)
Citation Context ...te as min XI:I∈I � I, J ∈ C E(XI, XJ) < τ (f(xI, xJ; ξ) − E(XI, XJ)) 2 (43) Standard optimization techniques can be used to solve this problem. In our prototype we use the simplex method des=-=cribed in [11]-=-. 6.2 Query Operator The query operator solves the dual problem of deriving the metric of the query space from the examples that the user gives in the display space. Let us assume that the user has se... |

1409 |
Features of Similarity
- Tversky
(Show Context)
Citation Context ...(x)dx = 1 is a weighting function that we will use in the following to manipulate the distance function. Note that with this definition the distance is not symmetric. For a justification of this, see =-=[19, 23]-=-. This distance can be used to measure dissimilarities between images. An important property of this class of distances is given by the following theorem (proved in [15]): Theorem 5.1 Let H be a subgr... |

961 |
Texture Features for Browsing and Retrieval of Image data
- Manjunath, Ma
- 1996
(Show Context)
Citation Context ...when applied to the feature space within a certain (usually quite narrow) range of semantic significance of these features. Extracting features like color histograms or Gabor descriptors for textures =-=[9]-=- may satisfy this requirement for certain classes of images. A second sense in which the requirement can be understood is that of coverage. In this case we require that the feature space support many ... |

362 |
A Theory of Semiotics
- Eco
- 1976
(Show Context)
Citation Context ...es that, if a record is produced by a rule like j : R → α1R1α2 · · · αnRnαn+1 1 It has been recognized that this behavioristic interpretation impoverishes somehow the semiotic space of the im=-=age-sign [2], but ev-=-en this impoverished semiotic space is sufficient for our purposes. 2 In this paper, we will commit the slight imprecision of using the terms “meaning” and “semantics” interchangeably. 2 (2)sw... |

312 | Fast multiresolution image querying
- Jacobs, Finkelstein, et al.
- 1995
(Show Context)
Citation Context ...d, and it not clear how to define the inner product 〈f, ψg〉. A possibility, often used in literature, is to treat the color image as three separate grayscale images, and to process them independe=-=ntly [6]-=-. Another possibility, introduced in [16], is to define an algebra of colors. In this case the values of Tf (g) in (30) are also colors. It is possible to extend the same definition to discrete groups... |

235 | Similarity measures
- Santini, Jain
- 1999
(Show Context)
Citation Context ...(x)dx = 1 is a weighting function that we will use in the following to manipulate the distance function. Note that with this definition the distance is not symmetric. For a justification of this, see =-=[19, 23]-=-. This distance can be used to measure dissimilarities between images. An important property of this class of distances is given by the following theorem (proved in [15]): Theorem 5.1 Let H be a subgr... |

111 | Dimensionality reduction for similarity searching in dynamic databases
- Kanth, Agrawal, et al.
- 1998
(Show Context)
Citation Context ...rs, as they come out of the image analysis algorithms. In many cases, we need to adjust these vectors for the convenience of the database operations. A very common example is dimensionality reduction =-=[12]-=-. In this case, we will say that we obtain a (reduced-dimensional) stored view of the feature space. The operators defined on the feature space are used for this purpose. The most common are: Projecti... |

58 |
Introduction to Lie groups and Lie algebras
- Sagle, Walde
- 1973
(Show Context)
Citation Context ...h a metric, this will naturally induce a metric in the space of transformed images. Consider the space L 2 (X, m), where m is a measure on X, and a Lie group G which acts freely and homogeneously on X=-=[14]. We -=-will assume that G, as a manifold, is endowed with a 14sRiemann metric. If g ∈ G, then a representation of G in L 2 (X) is a homomorphism π : G → L 2 (X) (23) The irreducible unitary representati... |

22 |
Wavelets associated with representations of the affine Weyl-Heisenberg group
- Torresani
- 1991
(Show Context)
Citation Context ...nd scale, in addition to the three color dimensions R, G, B. In this case the feature space is diffeomorphic to IR 6 . In other applications, we generate the transform using the Weyl-Heisenberg group =-=[15, 22],-=- obtaining transformation kernels which are a generalization of the Gabor filters [7]. In this case, in addition to the six dimensions above we have the direction θ of the filters, and the feature sp... |

20 | n-dimensional affine Weyl-Heisenberg wavelets
- Kalisa, Torrésani
- 1993
(Show Context)
Citation Context ...e need to endow the space F of image functions with a suitable metric structure. It is a well known fact that groups of transformations can be used to generate continuous wavelet transforms of images =-=[7]-=-. If the transformation group from which we start is endowed with a metric, this will naturally induce a metric in the space of transformed images. Consider the space L 2 (X, m), where m is a measure ... |

18 |
Text Databases and Information Retrieval
- Riloff, Hollaar
- 1996
(Show Context)
Citation Context .... Suppose a user enters a set of words A. It is possible to define the distance from the keyword set A to the textual portion of a visual concept label χ using normal information retrieval techniques=-= [13]. L-=-et d(A, χ) be such a distance. Similarly, it is possible to determine the distance between two concepts d(χ1, χ2). Then the textual query A can be transformed in a configuration of the display spac... |

17 |
Exploratory Image Databases
- Santini, Gibson
- 2001
(Show Context)
Citation Context ... Another possibility, introduced in [16], is to define an algebra of colors. In this case the values of Tf (g) in (30) are also colors. It is possible to extend the same definition to discrete groups =-=[15] by cons-=-ider a discrete subgroup GD ⊆ G. If G is a Lie group of dimension n, consider a set of indices I ⊂ Z n . We consider the set GD = {gj ∈ G : j ∈ I} , (26) under the usual closeness conditions t... |

12 | User interfaces for emergent semantics in image databases
- Santini, Jain
- 1999
(Show Context)
Citation Context ... this level it is no longer possible to continue the subdivision, and the complete semantic characterization of images goes beyond the simple enumeration of the objects in it and their relations (see =-=[20]-=- for examples). Semantic level beyond the objects are used very often in evocative scenarios, like art and advertising [1]. There is, for instance, a fairly complex theory of the semantics associated ... |

7 |
The Art of Color, Reinhold Pub
- Itten
- 1961
(Show Context)
Citation Context ...es). Semantic level beyond the objects are used very often in evocative scenarios, like art and advertising [1]. There is, for instance, a fairly complex theory of the semantics associated with color =-=[5]-=-, and with artistic representational conventions [3]. The full meaning of an image depends not only on the image data, but on a complex of cultural and social conventions in use at the time and locati... |

6 |
Art and Illusion. A study in the pshychology of pictorial representation
- Gombrich
- 1965
(Show Context)
Citation Context ... often in evocative scenarios, like art and advertising [1]. There is, for instance, a fairly complex theory of the semantics associated with color [5], and with artistic representational conventions =-=[3]-=-. The full meaning of an image depends not only on the image data, but on a complex of cultural and social conventions in use at the time and location of the query, as well as on other contigiencies o... |

5 |
A parametric class of discrete Gabor expansions, preprint
- Li, Healy
- 1995
(Show Context)
Citation Context ..., this transform has a metric structure just like the continuous transform. Note that the functions ψj = πD(gj)ψ will not in general form an orthogonal basis of L 2 (IR 2 ), but an overcomplete fra=-=me [8]-=-. An image can be completely represented by the set of coefficients of its transform. As we will see in the following section, this representation is very convenient for defining metrics in the image ... |

5 | Evaluation vademecum for visual information systems
- Santini
- 2000
(Show Context)
Citation Context ...expect. Evaluation of a highly interactive system like El Niño is a complex task, which could be the subject of a paper of its own. The interested reader can find some general ideas on the subject in=-= [17]-=-. 8 Conclusions In this paper we have defined a new model of interface for image databases. The motivation for the introduction of this model comes from an analysis of the semantics of images in the c... |

5 | The El Niño’; Image Database System
- Santini, Jain
- 1999
(Show Context)
Citation Context ...itable graphic front-end (see Fig. 5), whose characteristics vary depending on the particular application. Later in this paper we will consider the graphical front-end of our database system El Niño =-=([18]-=-). In this section we will be concerned with the operators that mediate the exploration of the database. 4.1 Operators in the Feature Space A feature is an attribute obtained applying some image analy... |

5 |
Regularization of incorrectly posed problems
- Tihonov
- 1963
(Show Context)
Citation Context ...egative effect is still an open problem. If we have identified a function ∆F (we will return to this problem shortly), then the metric optimization problem can be regularized (in the sense of Tihono=-=v [21]) as E = min -=-ξ � (f(xI, xJ; ξ) − qIJ) 2 + w∆F(f(·, ·; ξ), N) (48) I,J∈D that is, we try to find a solution that departs as little as possible from the natural distance in the feature space. For the de... |

4 |
Computer analysis of TV spots: The semiotics perspective
- Caliani, Pala, et al.
- 1998
(Show Context)
Citation Context ... beyond the simple enumeration of the objects in it and their relations (see [20] for examples). Semantic level beyond the objects are used very often in evocative scenarios, like art and advertising =-=[1]-=-. There is, for instance, a fairly complex theory of the semantics associated with color [5], and with artistic representational conventions [3]. The full meaning of an image depends not only on the i... |

3 |
The CASE philosophy
- Gibson
- 1989
(Show Context)
Citation Context ...ngs of its components. In other words, the meaning of a record is compositional. The meaning of a stimulus in a given situation is related to the set of possible actions of an actor in that situation =-=[10]-=-. In a database situation, the only possible actions are asking queries and answering them. Then, if Q is the set of all possible queries, the meaning of a record, or a fragment of a record R, can be ... |

1 |
search of information in visual media
- In
- 1997
(Show Context)
Citation Context ...tion we describe in greater details the principles behind the design of exploratory interfaces. 4 Exploration Operators The exploration interface is composed of three spaces and a number of operators =-=[4]. -=-The operators can be transformations of a space onto itself or from one space to another. The three space on which the interface is based are: • The Feature space F. This is the space of the coeffic... |

1 | Distance preservation in color image transforms - Santini - 2000 |