Image analysis – Image transformation or preprocessing – Image storage or retrieval
Reexamination Certificate
2001-09-21
2004-10-12
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image transformation or preprocessing
Image storage or retrieval
C382S190000, C382S209000, C382S278000, C358S403000, C707S793000, C345S960000
Reexamination Certificate
active
06804420
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to an information retrieving system with which the user specifies a plurality of pieces of information that are similar to desired information are indicated as search keys, a retrieving operation is carried out based on the feature of the plurality of selected retrieval keys, and the retrieval results are displayed on a display screen so that information pieces resembling each other are arranged closer to each other, so that the user is allowed to find desired information quickly.
2. Related Background Art
According to a conventional information retrieving system for retrieving desired information from among massive amount of information, a retrieval condition and information as retrieval targets are compared, and the retrieval results are displayed in a list in which they are arranged in a matching degree decrement order. In this case, without an appropriate retrieval condition being given, retrieval results with high precision cannot be obtained. In the case where image information is retrieval targets, for instance, it is difficult for the user to find and specify an appropriate retrieval request, so that many users input vague retrieval conditions. This causes a lot of retrieval results to be obtained, thereby the user have to check a lot of contents of the retrieval results visually for finding the desired information, which requires enormous labor and time.
As a technique of selecting the retrieval keys, JP2000-250943A discloses a technique for making the retrieving operation efficient by arranging pieces of information as retrieval results visually in a virtual three-dimensional space according to feature vector indicative of features of the respective retrieved information pieces. In this technique, one retrieval key selected by the user from information displayed on a display screen is used as a retrieval request. In this case, the retrieval key and information as the retrieval targets are expressed as feature vector, and degrees of resemblance between pieces of information are calculated according to differences between the elements of feature vectors, so that matching degrees are determined. Here, elements of the feature vector are treated equivalently.
Furthermore, JP2000-29885A discloses a technique for improving the efficiency in browsing the information, in which, while two or three elements of feature vector are switched at predetermined timing, information groups are displayed at once in a two-dimensional or three-dimensional space formed by axes according to the selected elements of feature vector.
When the user selects retrieval keys that are similar to the desired information, the user often brings only a part of elements of the feature vector into focus. Therefore, to retrieve information that the user truly requests with high precision, it is necessary to estimate precisely the elements of the feature vector that the user brings into focus, and then, to carry out a retrieving operation with only the elements of feature vector thus attracting the attention of the user. The technique disclosed by JP2000-250943A is unable to select appropriate retrieval keys, since therein the elements of the feature vector are treated equivalently.
Furthermore, in the case where various types of feature vector are used, likewise, it is necessary to carry out a retrieving operation by using only the feature vector that the user brings into focus among others. On the other hand, JP11(1999)-238078A discloses a technique in which, in the case where various types of feature vector are used, differences between feature vectors of similar types among the various types of feature vectors are determined, and elements of a feature vector with smaller differences are used in a retrieving operation. Here, in order to improve the browsability of retrieved results and to improve the efficiency in looking through the same, it is necessary to locate the similar ones among pieces of information more closely, and further, it is preferable that pieces of information are dispersed appropriately over the retrieval result display screen. With higher dispersion, blocks of information having common features tend to be formed, thereby allowing the efficiency in looking through information to be improved. However, according to the technique disclosed by JP11(1999)-238078A, a technique for selecting elements of feature vector used in retrieval is disclosed, whereas nothing is mentioned about a method for selecting elements of feature vector used in arrangement of the retrieval results for showing the retrieval results to the user. In the case where the elements of feature vector used for the retrieval are selected for arrangement, the retrieval results cannot be dispersed appropriately since they resemble the vector values. This problem also occurs to the selection of elements from a feature vector in vector representation in the case where only one feature value is used.
Furthermore, in the aforementioned technique of JP2000-29885A, alternatively, it is attempted to cope with the problem of feature vector selection by switching the elements of feature vector used for the space axes as time elapses in the arrangement display. However, the switching of the feature vector with elapse of time is irrelevant to the user's intention, and there arises a drawback in that the user has to wait until an intended element of feature vector is selected. Therefore, retrieval fit for the user's intention is not necessarily carried out.
SUMMARY OF THE INVENTION
Therefore, to solve the foregoing problems, it is an object of the present invention to generate a retrieval result displaying screen fit for the user's intention of retrieval and to enable efficient retrieval, by, in retrieval, appropriately selecting information to be used in retrieval according to a plurality of retrieval keys selected by the user, executing retrieval using the selected information, and further, in arrangement of the retrieval results, by appropriately selecting information to be used in arrangement according to the selected retrieval keys.
To achieve the foregoing object, an information retrieving system of the present invention includes: a retrieval key indicating section with which a user indicates a plurality of retrieval keys; a feature vector extracting section for extracting elements of feature vector that correspond to the retrieval keys, respectively; a retrieval feature vector selecting section that compares feature vector elements and calculates averages of distances between corresponding feature vector elements, and selects a prescribed number of feature vector elements in a distance average ascending order as retrieval feature vector to be used in retrieval; a retrieving section that retrieves information stored in an information storing section by using the retrieval feature vector as retrieval executing keys; an arrangement feature vector selecting section that selects the elements of the feature vector that have not been extracted as the retrieval feature vector as arrangement feature vector among the elements of the feature vector vector of each retrieval result; an arranging section that classifies and arranges results of the retrieval by using the arrangement feature vector as features of the arrangement; and a display section that displays a result of the arrangement.
With the foregoing configuration, since only the feature vector elements that the user brings into focus, among the elements of feature vector, are used in retrieval, retrieval with high precision fit for the user's intention is enabled. Further, since the feature vector elements other than the feature vector element used in the retrieval are used in arrangement of the retrieval results, the retrieval results can be dispersed appropriately, whereby the retrieval results can be displayed in an arrangement excelling in browsability.
It should be noted that the arrangement feature vector selecting section selects, as arrangement feature vector, a
Endo Susumu
Masumoto Daiki
Sashida Naoki
Shiitani Shuichi
Uehara Yusuke
Kassa Yosef
Mehta Bhavesh M.
LandOfFree
Information retrieving system and method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Information retrieving system and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Information retrieving system and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3323123