Data processing: database and file management or data structures – Database design – Data structure types
Reexamination Certificate
2000-12-01
2004-08-24
Rones, Charles (Department: 2175)
Data processing: database and file management or data structures
Database design
Data structure types
C707S793000, C707S793000, C709S202000
Reexamination Certificate
active
06782395
ABSTRACT:
TECHNICAL FIELD
The present invention concerns, in accordance with a first aspect, a method of seeking images, using an example image, from amongst a plurality of images stored in a database, each of the stored images being associated with a data item representing at least one characteristic of the visual content of the image. The invention also concerns a device able to implement such a method.
According to a second aspect, the present invention concerns a method of indexing digital images taking into account the location of regions of interest in the images. The invention also concerns a method of seeking images from an example image, from amongst a plurality of images stored in a database, the images being indexed according to the above-mentioned indexing method. The invention also concerns a device able to implement such methods of indexing and seeking images.
BACKGROUND OF THE INVENTION
A conventional method of seeking digital images in a database containing images is based on a system of indexing the images in the database.
The purpose of the indexing system is to associate, with each image in the base, an item of information characteristic of the content of the image referred to as the index of the image. All these information items form the index of the database.
A user can then interrogate the image database through a request containing an item of information characteristic of the type of image sought. The content of the request is then compared in accordance with a search strategy with the content of the index of the database.
Finally, the image in the base whose indexed information item has the greatest similarity to the content of the request is then extracted. A plurality of images extracted from the base can also be presented to the user, ordered according to their degree of similarity to the request.
In a traditional system of indexing digital images, the index of the database is composed of textual descriptors of the images stored.
The request of the user then consists of key words describing the characteristics of the content of the image to be sought.
This type of indexing by textual descriptors has the drawback of being imprecise, in particular because the same image may be described in different ways by different users.
In order to mitigate this type of drawback, the need has therefore been felt to develop techniques for representing and extracting the semantic content of a digital image.
Methods have appeared in which an image is characterised according to the distribution of the colours or textures making it up.
In other methods, an image is characterised by the shape or contour of an object making it up.
However, all these descriptors, referred to as “primitives”, of the image, reflect only physical characteristics of the image, and are therefore of a low semantic level.
In order to increase the semantic character of the indexing of the images, indexing systems which use a combination of low-level primitives are beginning to appear.
One of the most well-known is certainly the QBIC (“query-by-image-content”) system developed by IBM.
To obtain more details on this system, reference can be made to the article “
The QBIC Project; Querying Images by Content Using Color, Texture, and Shape”
, by Niblack, W et al., which appeared in IBM Computer Sciences Research Report, pp 1-20 (Feb. 1, 1993). Reference can also be made to U.S. Pat. No. 5,579,471 of IBM entitled “
Image query system and method”.
The QBIC system makes it possible to find digital fixed (or video) images from a request by means of the example. In the particular case of fixed images, this request is defined either as a complete image or as an object with a rectangular or arbitrary shape, extracted from an image or defined by the user.
The content of the images in the database is characterised by its colour distribution (histogram), texture and shape.
Where the request is defined as a portion, also referred to as a region, of an image, the similarity measurement also takes account of the spatial position of this region.
However, in the QBIC system and in the indexing systems of this type, the indexing of an image stored in the database relates to the entire content of the image, without consideration of the zone or region which particularly represents the visual content of the image.
Moreover, the request at best relates only to a single portion of the example image.
However, in a given image, several distinct regions may contain a great deal of semantic information. Consequently, in spite of the combination of several primitives for indexing the content of the image, this type of index is not sufficiently representative of the semantic content of the image.
Thus, in the indexing systems of the QBIC system type, an analysis of the similarity between an index derived from a request and that derived from a stored image is normally still insufficiently precise.
In the following description, the expression “region of interest” (ROI) is used to mean a delimited portion of a given image predefined as containing elements characteristic of the visual content of the image, that is to say as containing a great deal of semantic information.
The present invention aims to remedy the aforementioned drawbacks of the search and indexing systems of the prior art.
SUMMARY OF THE INVENTION
To this end, in accordance with a first aspect, the present invention concerns a method of seeking images, from an example image, from amongst a plurality of images stored in a database, each of the stored images being associated with a data item of a first type representing at least one characteristic of the visual content of the image. In accordance with the invention this image search method includes the following steps:
each time a new image is stored in the database, associating with the new image a data item of a second type representing the location of at least one region of interest in the new image;
receiving a data item of the second type representing the location of at least one region of interest in the example image;
obtaining a data item of the first type associated with the example image;
calculating a similarity between the example image and each of the images amongst at least one subset of said stored images, the similarity being calculated from the data of the first type and of the second type associated respectively with the example image and with the stored image; and
supplying at least one image, referred to as the result image, of the database, the result image being selected from amongst the stored images in the database according to its degree of similarity with the example image.
In this way, taking account of the data representing the location of at least one region of interest (data item of the second type), in the example image and in the images of the database, the search for images in the database is more precise. This is because on the one hand, for each image in the database, the database manager has taken care to isolate the semantic content of the image through this data item of the second type. The user, on the other hand, through the data item of the second type associated with the example image, can also specify his request by choosing in the image certain regions which correspond to the one which he wishes to find. Consequently there is obtained a similarity closer to reality compared with the one obtained when this type of data is not taken into account both in the example image and in the images in the database.
The present invention also concerns a device for seeking images, from an example image, from amongst a plurality of images stored in a database. In accordance with the invention this device has means adapted to implement an image search method as succinctly defined above.
The present invention also concerns a computer, comprising such an image search device or means adapted to implement the above-defined image search method. The invention also relates to a computer program including one or more sequences of instructions able to implement the above-defined image search method when th
Canon Kabushiki Kaisha
Fitzpatrick ,Cella, Harper & Scinto
Mahmoudi Hassan
Rones Charles
LandOfFree
Method and devices for indexing and seeking digital images... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and devices for indexing and seeking digital images..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and devices for indexing and seeking digital images... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3290777