WO2000039712A9 - Improved techniques for spatial representation of data and browsing based on similarity - Google Patents
Improved techniques for spatial representation of data and browsing based on similarityInfo
- Publication number
- WO2000039712A9 WO2000039712A9 PCT/US1999/030298 US9930298W WO0039712A9 WO 2000039712 A9 WO2000039712 A9 WO 2000039712A9 US 9930298 W US9930298 W US 9930298W WO 0039712 A9 WO0039712 A9 WO 0039712A9
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- node
- subnode
- control points
- control point
- instructions
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2137—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps
- G06F18/21375—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps involving differential geometry, e.g. embedding of pattern manifold
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/912—Applications of a database
- Y10S707/913—Multimedia
- Y10S707/915—Image
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/912—Applications of a database
- Y10S707/918—Location
- Y10S707/921—Spatial
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/953—Organization of data
- Y10S707/955—Object-oriented
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/953—Organization of data
- Y10S707/956—Hierarchical
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/953—Organization of data
- Y10S707/957—Multidimensional
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99931—Database or file accessing
- Y10S707/99933—Query processing, i.e. searching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99931—Database or file accessing
- Y10S707/99933—Query processing, i.e. searching
- Y10S707/99936—Pattern matching access
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99931—Database or file accessing
- Y10S707/99937—Sorting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99942—Manipulating data structure, e.g. compression, compaction, compilation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99943—Generating database or data structure, e.g. via user interface
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99944—Object-oriented database structure
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99944—Object-oriented database structure
- Y10S707/99945—Object-oriented database structure processing
Definitions
- the present invention relates generally to browsing and database technology and, more particularly, to image browsing and image database technology.
- MDS MultiDimensional Scaling
- the relative location of the nodes is dependent upon the object similarities or dissimilarities, which are interpreted as a set of distances between the nodes.
- the object similarities or dissimilarities can be determined by a variety of techniques, which can then be used to determine the set of distances between the nodes in the MDS space.
- MDS is a computationally expensive technique.
- MDS can be impractical due to its global nature, which requires extensive matrix processing.
- the typical MDS techniques may not be practical for larger image databases (e.g., on the order of hundreds or thousands of images) .
- the typical MDS techniques do not necessarily provide biologically plausible techniques for the spatial representation of data, and in particular, do not allow for intuitive browsing of, for example, images in an image database.
- a process for querying a computer-implemented hierarchical MultiDimensional Scaling (MDS) database for images includes measuring dissimilarity of a set of images using feature detectors; obtaining a set of distances between control points corresponding to images in a root node; performing a single node update at the root node to determine a first position in the root node of an image being queried or added; determining a first bounding box for a first subnode, in which the first subnode is a child of the root node; and determining a list of traversed nodes and traversed control points, performing a single node update at the first subnode, and sorting distances to the traversed control points in the traversed nodes, in which the first sub
- the process further includes obtaining a list of images in a second subnode, in which the second subnode is the child of the first subnode; and repeating the performing of the single node update and the determining of a second bounding box for the second subnode.
- a process for a computer- implemented hierarchical spatial database of objects includes determining distances between control points corresponding to objects in a root node of the hierarchical spatial database of objects; and determining a position of a first control point in the root node for a first object, in which the first object is being queried, and in which the hierarchical spatial database of objects includes the root node and a first subnode, the first subnode being a child of the root node.
- the process can further include traversing a first subnode and performing a single node update on the first subnode; performing the single node update at a leaf node, the leaf node being a descendant of the first subnode; and determining traversed subnodes and traversed control points, and sorting distances between the traversed control points in the traversed subnodes and the control points in the root node to the control point for the first object.
- the process can further include adding the first object to the hierarchical spatial database of objects, in which the leaf node is subdivided if the leaf node is full, and in which multidimensional scaling is executed on the leaf node and updating all bounding boxes in the traversed path to the first object.
- the hierarchical spatial database of objects can be initialized by executing instructions for approximating a convex hull. For example, this process can be used for browsing and modifying a hierarchical MDS database for images, in which the images are stored on one or more memories (e.g., local or remote memories of data processing devices) .
- a process for a computer- implemented hierarchical spatial database of objects includes calculating multiple stress vectors, in which the multiple stress vectors represent stress factors between a first control point and multiple control points of the hierarchical spatial database of objects, and in which the multiple control points correspond to multiple objects, and the first control point corresponds to an object being queried; and mapping the multiple stress vectors to multiple deformation vectors; combining the multiple deformation vectors into a single node update vector; and updating the first control point by moving a position of the first control point based on a fraction of the single node update vector.
- the multiple control points can include multiple source control points, and the first control point can represent a target control point, in which the calculating of the multiple stress vectors includes the following: storing values for multiple source bundle fields and multiple target bundle fields; and determining multiple source field values, the multiple source field values corresponding to the multiple source control points, the multiple source control points in a neighborhood of the target control point, in which a position of the target control point is modified using the source field values, and in which the stress on the target control point in a node of the hierarchical spatial database of objects is minimized.
- the fields can advantageously correspond to local fields (e.g., as opposed to the global stress factor of standard MDS techniques) or anisotropic fields.
- FIG. 1 is a block diagram of a data processing system in accordance with one embodiment of the present invention.
- Figure 2 is a block diagram of the various program modules and a hierarchical spatial database for images stored in the memory of Figure 1 in accordance with one embodiment of the present invention.
- Figure 3 is a flow diagram of an initialization of the hierarchical spatial database of Figure 2 in accordance with one embodiment of the present invention.
- Figure 4 is a flow diagram of a query and an add performed on the hierarchical spatial database of Figure 2 in accordance with one embodiment of the present invention.
- Figure 5 is a flow diagram of a single node update of the hierarchical spatial database of Figure 2 in accordance with one embodiment of the present invention.
- Figure 6 is a flow diagram of a biologically plausible implementation of the hierarchical spatial database of Figure 2 that allows for more intuitive browsing of the images in accordance with one embodiment of the present invention.
- Figure 1 illustrates a data processing system in accordance with one embodiment of the present invention.
- Figure 1 shows a computer 100, which includes three major elements.
- Computer 100 includes an input/output (I/O) circuit 120, which is used to communicate information in appropriately structured form to and from other portions of computer 100 and other devices or networks external to computer 100.
- Computer 100 includes a central processing unit (CPU) 130 (e.g., a microprocessor) in communication with I/O circuit 120 and a memory 140 (e.g., volatile and nonvolatile memory) .
- CPU central processing unit
- memory 140 e.g., volatile and nonvolatile memory
- a raster display monitor 160 is shown in communication with I/O circuit 120 and issued to display images (e.g., video sequences) generated by CPU 130. Any well-known type of cathode ray tube (CRT) display or other type of display can be used as display 160.
- a conventional keyboard 150 is also shown in communication with I/O circuit 120.
- computer 100 can be part of a larger system.
- computer 100 can also be in communication with a network, such as connected to a local area network (LAN) or the Internet .
- LAN local area network
- computer 100 can include circuitry that implements improved techniques for spatial representation of data and browsing based on similarity in accordance with the teachings of the present invention.
- the present invention can be implemented in software executed by computer 100 (e.g., the software can be stored in memory 140 and executed on CPU 130) , as further discussed below.
- the present invention can also be implemented in circuitry, software, or any combination thereof for various other types of data processing devices.
- the present invention can be implemented in a digital camera to provide for browsing and efficient storage of digital images stored in a spatial representation in a memory of the digital camera (e.g., in a local disc or removable memory, such as a floppy disc or flash memory card) .
- MDS techniques for spatial representation of data based on similarity or dissimilarity are computationally expensive.
- spatial representation of image data in an MDS space does not account for local effects that adding an image can have on the relative location of other nearby images in the neighborhood of the added image.
- MDS techniques typically account for the effect of an added image using a global factor (e.g., a global factor of stress) rather than a local factor.
- a global factor e.g., a global factor of stress
- a technique for a computationally efficient spatial representation of images in a database and intuitive browsing of the stored images based on similarity is provided.
- the technique advantageously utilizes a hierarchical MDS technique (e.g., a tree-based hierarchy) for computational efficiency (e.g., a technique for accessing a hierarchical MDS database for images that is of order less than O(N), such as 0(log(N))).
- a hierarchical MDS technique e.g., a tree-based hierarchy
- the technique implements an MDS database that can account for local effects, which facilitates intuitive browsing of the images.
- a hierarchical MDS database is provided. More specifically, a hierarchical MDS database is provided by mapping data, such as image data, in a space referred to as a manifold.
- a manifold in general, is a space X together with a set of homeomorphisms ⁇ #>_ . ⁇ , ⁇ _ . : X ->5R n , such that for each xe X, ⁇ ⁇ (x) is defined for some i.
- n is the dimension of the manifold, and X inherits metric space topology from
- ⁇ t is expected to be more than a homeo orphism, and different types of manifolds can be defined by describing how the charts ⁇ i interact where they overlap. For instance, a differentiate manifold is a manifold such that ⁇ t ° ⁇ >j _1 is a diffeomorphism for all pairs (i,j).
- a manifold can be used to construct a hierarchical MDS space, which is flexible and nonlinear.
- an MDS space for images can be implemented as a manifold, and interactions between charts compatible with MDS can be defined.
- global mapping created by MDS may not be provided in this approach, the ability to describe a space in which feature descriptions change with location is provided.
- an improvement in computational efficiency due to the hierarchical structure of the MDS database is provided, as described below.
- a configuration represents a set of points together with a set of labels for the points.
- two configurations can describe the same objects if there are two sets of points that share the same labels.
- a proximity is assigned, and a mapping from the set of proximities to a set of distances within the embedding space is provided.
- Two configurations have non-empty intersection or overlap if they share objects and, therefore, their labels and proximities.
- the labels are associated with objects, whereas the points provide a particular representation of the objects.
- the proximities are likewise attributed to the objects, whereas the distances are related to a particular representation.
- a set of objects can have more than one representation, and these representations are the configurations.
- the objects represent control points determining the shape (e.g., deformation) of the image space. Therefore, a nonlinear manifold is described, which is at any point a continuous map of 5R n .
- the twists and turns of the manifold can usually be described by the description of a discrete set of control points.
- These control points are embedded in neighborhoods (e.g., a particular node in the hierarchical MDS space corresponds to a particular neighborhood) , which are copies of open sets in 9? n , and mappings between such neighborhoods are defined by their actions on the control points.
- a relaxation transformation of one configuration into another configuration of the same object is a Procrustes transformation composed with an MDS optimization.
- the Procrustes transformation of the first configuration lies in the basin of attraction of the second configuration.
- An MDS manifold is a manifold, together with a set of configurations, such that to each configuration there corresponds a chart ( ⁇ ,U) of the manifold, with the property that if two such configurations and charts overlap, then the two configurations are elastic deformations of each other.
- MDS transformations can be represented as deformations of the MDS manifold itself, which minimize stress in the MDS manifold.
- Stress for MDS can be represented as some (possibly normalized) cost function that compares the distances into which the proximities are mapped with the distances in the configuration.
- stress can be determined based on the action of each of the control points on each of the other control points.
- each (control) point is viewed as creating a field that induces a force on the other points, and thus, causes a stress at the other points.
- S n_1 x B together with the original manifold B, and a mapping ⁇ , taking any element in (s,b) 6 E to the point b on B.
- a sphere bundle over a manifold is a space with a sphere attached to each (control) point in the MDS space.
- a large collection of vector fields can be represented as a scalar field on E, together with the assignment of an angle (i.e., direction) to each point b.
- the vector at b is then the vector having direction assigned by the angle, and length assigned by the scalar corresponding to it on E.
- a vector backprojection of a sphere bundle over a set of points is a function from the fibers of the bundle to values on the fibers over the bundle. In other words, it takes all the values on the sphere over a point and calculates a single vector in the bundle at that point (i.e., calculates a single real number and a direction) .
- the action of a point in an MDS manifold on another point of the MDS manifold is a value on the fiber over the acted on point .
- the value is added to other actions that have values at the same position on the fiber.
- the vector stress at a point x in an MDS manifold is a vector backprojection of the accumulated actions at x due to the other points within a neighborhood of x.
- This approach allows for an implementation of MDS using local effects of stress, as further discussed below with respect to Figure 6.
- One form of vector stress is caused by real deformations. Stress caused by real deformations can be implemented using the technique for a single node update, as further discussed below with respect to Figure 5. In this case, the stress is calculable by a discrepancy between the mapping of proximities into distances and the configuration.
- a hierarchical spatial database and various program modules employing these techniques are described below with respect to Figure 2.
- FIG. 2 is a block diagram of the various program modules and a hierarchical spatial database for images stored in the memory of Figure 1 in accordance with one embodiment of the present invention.
- the program modules can be implemented in a variety of programming languages such as JAVA, C++, or any other programming language, or any combination of programming languages, and executed on CPU 130.
- Figure 2 illustrates various program modules and an image database stored in memory 140.
- a Graphical User Interface (GUI) /control module 200 is in communication with a file manager 210, a query manager 250, and a feature detector and scorer manager 270.
- GUI Graphical User Interface
- various program modules and databases can communicate via message passing.
- Feature detector and scorer manager 270 is used to compute scores for an image to be queried or added relative to other images in a given node. Various feature detectors and scoring techniques can be used as would be apparent to one of ordinary skill in the art.
- the feature detector (s) and scorer of feature detector and scorer manager 270 form distances to these points that are sent to a feature detector results 240.
- File manager 210 manages a registry 220 (e.g., maintained in a file or a database) , a configuration of hierarchical MDS space 230 (e.g. maintained in a file or a database) , and feature detector results 240 (e.g. maintained in a file or a database) .
- Registry 220 stores locations of the images, such as the locations of image files stored in memory 140 or in another local or remote memory, such as a URL for a World Wide Web site location accessed via the Internet.
- Registry 220 also stores locations of the feature detectors that can be loaded to interpret the data stored in feature detector results 240.
- Configuration of hierarchical MDS space 230 includes configuration data for reconstruction of the hierarchical MDS space (that was previously constructed) .
- Feature detector results 240 stores the results of the feature detector analyses for each image represented in the hierarchical MDS space. For example, if a color histogram feature detector and a wavelet feature detector are applied to an image, then feature detector results database 240 stores a set of histogram and a set of wavelet coefficients for the image.
- Query Manager 250 interacts with and modifies the hierarchical MDS space for querying and adding images. For example, query manager 250 executes various other program modules and dynamic library links to perform various operations including initializing, modifying, and querying the hierarchical MDS space, as further discussed below. A functional depiction of the hierarchical MDS space, as described above, is also illustrated in Figure 2.
- the hierarchical MDS space is implemented as a set of spaces corresponding to the above-described MDS manifolds.
- the hierarchical MDS space includes a root node 260.
- the root point set is returned via an OS-independent callback mechanism (e.g., implemented via try/catch blocks in C++).
- the next lower layer of nodes i.e., subnodes
- the next lower layer of nodes includes leaf nodes 272, 274, 276, and 278, and a node 268.
- the lowest layer of nodes includes a leaf 279, which is a child node (or leaf) of node 268.
- Figure 2 also illustrates an exploded view of node 266 and its leaf node 278.
- the (control) points control points are discussed above) in each of the exploded views of node 266 and leaf node 294 correspond to data objects, such as images.
- point 282 corresponds to a particular image
- the dashed box surrounding point 282 corresponds to its bounding box, which is the bounding box describing leaf node 278.
- the asterisk point 284 corresponds to an image that is being queried or is to be added, and as illustrated, it falls within bounding box 280.
- the image being queried/added is shown in the exploded view of leaf node 278 as asterisk point 294.
- root node 260 includes about 20 points, and the subnodes include fewer than 20 points.
- a leaf node may only include a few points.
- Figure 3 is a flow diagram of an initialization of the hierarchical spatial database of Figure 2 in accordance with one embodiment of the present invention.
- Figure 3 provides a technique for determining points in a hierarchical MDS space that approximate a convex hull (or at least ensure that the selected points are not merely a clump of points that are relatively near each other) .
- the hierarchical image space of Figure 2 is created by initial input of a starting set of distances (e.g., between images) .
- any node or subnode i.e., top layer or lower layers of the hierarchical space
- the root node i.e., top layer
- image distances e.g., about 20 images can be represented in the root node
- the root node approximates the convex hull of these distances by the technique illustrated in the flow diagram of Figure 3 and described below.
- the largest distance is selected from the set of distances, and the two points to which it corresponds are recorded. These two points must be on the convex hull, because if these two points are interior to any simplex constructed from convex hull points, a contradiction via the triangle inequality results.
- the next largest distance from the remaining 1 set of distances is selected. Either this distance has one or no points in common with the first, and add one or two points to the set. All distances are arranged in a triangle, representing the lower triangular portion of a matrix of distances between the selected points. In particular, the triangle is arranged such that each row of the triangle sums to a value greater than the row above, and the next largest distance is selected to add points.
- stage 306 whether the sum of distances to the previously selected points is greater than the previous sum is determined. If so, operation proceeds to stage 308. Otherwise, operation returns to stage 304. Referring to stage 308, whether sufficient points have been obtained to use for the parent node of the subdivide is determined. If so, operation proceeds to stage 310. Otherwise, operation proceeds to stage 312. At stage 312, if there are remaining points to try, then stages 304 and 306 are repeated, and if there are not any remaining points to try, operation returns to stage 302.
- This initialization technique can advantageously be implemented without any full sorts by pushing all the distances onto a heap to start. As many points from the heap as are needed can be drawn to complete the set .
- control points to determine a root configuration and bounding box.
- the distances to these points from the other points in the initial set are stored in configuration 230, and the remaining points are given starting configuration coordinates relative to this hierarchical MDS space by the single node update procedure, which is discussed below with respect to Figure 5.
- the selected points are positioned in the first node (root node) , and the positions of the points in the first node are stored in configuration 230.
- the first node is split into multiple nodes under the root, using a median cut technique. This splitting operation proceeds until the nodes are small enough or the set of nodes is full. In the former case, MDS is then run on the leaf nodes, and the initial MDS space is completed, and stored in configuration 230. In the latter case, the nodes that contain many members can be subdivided. Thus, this technique for the root node is recursively applied to the subnodes. It is important to notice that multiple bounding box descriptions are generated to effect the MDS manifold structure. For example, the root node retains its own bounding box, plus the bounding box in the root node coordinates of its children.
- each of these children likewise holds its own bounding box, and those of its children in its own coordinate chart.
- the bounding box that the root node holds describing child node i is not the same as the bounding box that i holds describing itself.
- a point in this space has a description in coordinates only with respect to a specific chart, which corresponds to a particular node in the hierarchy. Consequently a complete description of a particular set of coordinates includes an identifier for the node at which they are realized.
- the stages of operation of Figure 3 are implemented as program instructions stored in query manager 250 (e.g., or a program module or dynamic link library called by query manager 250) and executed on CPU 130.
- the properties (i.e., requirements) of an MDS manifold are satisfied by creating such a spatial hierarchy.
- this spatial hierarchy is consistent with the MDS manifold description provided above. More specifically, the MDS manifold description includes a chart ( ⁇ ,U) plus a set of coordinates
- FIG. 4 is a flow diagram of a query and an add performed on the hierarchical spatial database of Figure 2 in accordance with one embodiment of the present invention.
- a technique for accessing a hierarchical MDS database that is of order less than O(N) is now described.
- a large collection of images e.g., on the order of hundreds or even thousands (or more) stored images
- the dissimilarity of the collection of stored images is measured as distances using some feature set (e.g., the feature set may be application specific) , using feature detector and scorer manager 270.
- a list of images in the root node is obtained and sent to feature detectors and scorer manager 270 to obtain a list of distances between the images (control points) in the root node.
- a single node update (as described below with respect to Figure 5) is performed at the current node (e.g., root node or a subnode) , taking the coordinates found at the node above as initial conditions to the single node update at the current node.
- a bounding box around the found coordinates for the node at the next lower layer is determined. In other words, some of these images are "pushed" down to the nodes below the top layer by allowing their configuration at the top layer to be initial conditions for their positions in the lower layers of the hierarchical MDS database.
- next lower node is a leaf node is determined. If so, operation proceeds to stage 412. Otherwise, a list of images in the next lower node is obtained, and the list is sent to feature detectors and scorer manager 270 to obtain a list of distances between the images in that node, at stage 414, and stages 406 and 408 are then repeated.
- stage 412 lists of nodes and points of the traversed path are determined, and a single node update at the leaf node is executed.
- the distances to the points in the traversed nodes are sorted.
- the appropriate images can be displayed (e.g., output on monitor 160) in order of similarity based on the sorted list using Graphical User Interface (GUI) /control module 200. Accordingly, a user of the MDS system for images can browse the images based on the sorted results.
- GUI Graphical User Interface
- a user of the MDS system for images can browse the images based on the sorted results.
- stage 420 whether the leaf node is full is determined. If so, the leaf node is appropriately subdivided at stage 422.
- full MDS is executed on the leaf node to which the new image is being added using previously calculated coordinates for the image being added as the starting configuration, and all bounding boxes in the traversed path to the new control point are updated.
- a previous query for an image can be stored in memory (e.g., a flag is set) such that if the same query is requested again, the query operation described above need not be repeated (assuming that the hierarchical MDS database has not changed since the last query) .
- Figure 4 provides a less than O (N) implementation, and in particular, an 0(log(N)) implementation is provided for querying or adding an image to the hierarchical MDS database .
- the stages of operation of Figure 4 are implemented as program instructions stored in query module 250 (e.g., or a program module or a dynamic link library called by query module 250) , and executed on CPU 130.
- adding an image is done by first applying the above query technique, updating all necessary bounding boxes, and adding the point to the necessary leaf nodes. In the leaf nodes, this may or may not cause either a split, if there is space at the current node, or a subdivide, if there is not space at the current node .
- a single node update technique for reducing the computational complexity of updating single nodes in an MDS configuration is disclosed in co-pending U.S. Patent Application entitled, "METHOD AND APPARATUS FOR UPDATING A MULTIDIMENSIONAL SCALING DATABASE", to Hawley K. Rising III, filed October 20, 1998, Serial No. 09/176,052, Attorney Docket No. 50M2653, which is hereby incorporated by reference in its entirety.
- a vector sum is calculated at the point of the single node, via the formula where
- the vector calculated above is an expression of the deformation at the point * due to stress. This stress is calculable via the standard cost functions for MDS.
- the above vector quantity can advantageously be associated with the stress at the point xf by Hooke's law.
- it is possible to associate any quantity with the deformation perceived by the point x* at the point xf so long as there is a field that gives the mapping of vectors at X j to vectors at JC* .
- the field then becomes a mapping on the sphere bundle over the space (usually real Euclidean space) in which the MDS configuration is represented.
- the deformation is an inference, in which a perceived stress at JC* is mapped to an inferred deformation at ** .
- the field over the sphere bundle gives the logical rule for carrying out this inference.
- the set of inference data that are mapped onto the sphere over xf are combined in the single node update procedure by vector addition, but this is not required. Their combination constitutes a scalar or vector backprojection, and the rule used to combine them therefore determines the update taken.
- the above-described single node update technique can be used to provide properties to the metric space chosen.
- the metric space acts as a perfectly elastic continuum under the standard global stress model .
- the metric space can be modified to have any continuum properties under the above procedure, as further discussed below.
- Figure 5 is a flow diagram of a single node update of the hierarchical spatial representation of images of Figure 2 in accordance with one embodiment of the
- each target (destination) point lists source points that are within its range (for purposes of having stress effects on the target point) , and then from that list of points, it is determined whether the target point is within the range for each of the source fields of the points in the list.
- the field rule can be used to map this to a vector at x* representing deformation.
- the backprojection rule can be used to combine the vector deformations at xf into a single update vector.
- xf is updated by moving some fraction of the update vector.
- the stages of operation of Figure 5 are implemented as program instructions stored in query manager 250 (e.g., or in a program module or dynamic link library called by query manager 250) and executed on CPU 130.
- query manager 250 e.g., or in a program module or dynamic link library called by query manager 250
- the spatial representation of data in an MDS database should provide a representation of data that allows for intuitive browsing or searching of the stored data.
- an MDS model of image data it is desirable for an MDS model of image data to be consistent with (accurately,, model) human perception of the relationships among such image data.
- MDS can be described as a juxtaposition of a memory stage, which uses Self-Organizing Maps (SOM) , and a Radon transform method for detecting change, which has been accepted as a biologically plausible model.
- SOM Self-Organizing Maps
- Reichardt detectors can be constructed for many types of data and can be modified to use different types of tests for fit. When combined with the SOM stage, this becomes an adaptable form of spatial organization.
- the image control points are regarded as cluster centers, then vector data given by Radon transforms and Reichardt detectors provides the input necessary to generate the herein disclosed and improved MDS model.
- techniques for local fields and overlap patches for MDS-based databases are provided.
- a biologically plausible implementation of an MDS-based image database advantageously allows for more intuitive browsing of images.
- a bundle description of a generalized MDS technique is provided. This description splits the bundle description of MDS, which is discussed above, to allow for observer and observed, or prototype and new object distinctions, to be made in viewing an MDS generated space.
- adaptive field representations of user preferences, adaptive field representations of locality of similarity, adaptive field mappings that allow re- orientation, and modification of less than perfect feature detector output to fit recognition systems can be modeled in the MDS database.
- the technique for this embodiment is to generate a field at each control point, which represents the influence of the control point, and then to generate a field at the new control point . The interaction of these fields produces flexibility in the MDS space, as illustrated and discussed below with respect to Figure 6.
- Figure 6 is a flow diagram of a biologically plausible implementation of the hierarchical spatial database of Figure 2 that allows for more intuitive browsing of the stored images in accordance with one embodiment of the present invention.
- the MDS database using the pair of bundles allows for the following biologically plausible implementation of the MDS database.
- the fields for the source bundle and target bundle are determined as either formulaic expressions, storing any coefficients that are used to adapt a particular point, or as lookups, storing values that will be modified.
- the field values for all nearest neighbor points are determined (e.g., calculated or looked up in memory) .
- the field values that are not zero include a list of points to evaluate.
- the field value of the source point is determined.
- the distance between the source point and the target point is then modified by multiplicative application of the source and target field values.
- the position of the target point is modified in accordance with minimizing the stress on the target point.
- the stages of operation of Figure 6 are implemented as program instructions stored in query manager 250 (e.g., or in a program module or dynamic link library called by query manager 250) and executed on CPU 130.
- elasticity and viscosity in the context of MDS-related stress is also used to generate the local fields described above with respect to Figure 6.
- viscosity represents the ability of the system to react to stress at a distance
- elasticity corresponds to the division between adaptation and accommodation.
- elasticity corresponds to the relative amounts of accommodation/adaptation versus reaction/correction.
- elasticity can be used for techniques for developing adaptations to feature detector input in an MDS system. For example, models relying on extensions of fluid models of viscous stress and elastic stress simplify the creation of adaptive fields, because behavior can be predicted using models from fluid and continuum dynamics.
- modifying an MDS system using elasticity and viscosity advantageously allows for the modeling of multiple known effects in similarity experiments based on human perception.
- a variety of parameters can be ascertained by modeling multiple known effects in similarity experiments, such as accommodation, asymmetry, locality, and continuity.
- the modeling of an explicit elasticity at the target point allows explicit modeling of accommodation. This modeling can take the form of a formulaic value for the elastic strain, or just a readjustment of the proximity towards the derived distance.
- Locality can be modeled as follows: first, the influence of particular stored objects are of a limited range, and second, viscosity is modeled such that the size of the displacement of one part of the visual memory changes the range of objects over which memory is reorganized.
- an MDS system for image data can be implemented such that adding images containing a significant amount of the color blue does not effect (i.e., cause a reorganization of) any images containing mostly red, unlike an MDS image system that uses a global stress factor.
- continuity can also be modeled in an MDS system. For example, some differences between images are not monotonic in their differences in similarity.
- the MDS system in accordance with one embodiment of the present invention can account for such continuity issues by the shape and extent of the source and target fields.
- more computationally efficient techniques for accessing a hierarchical MDS database are provided.
- the techniques in accordance with the teachings of the present invention provide 0(log(N)) computational performance for accessing a hierarchical MDS database for images, which is very important for many practical implementations of an image database that can include hundreds (or even tens of thousands) of images.
- a variety of programming languages can be used to implement the techniques in accordance with the teachings of the present invention, such as the well- known C++ or JAVA programming languages, and a variety of operating system technology, file structure formats, and database technology can be utilized to implement the present invention.
- the present invention can also be implemented in hardware (e.g., an Application Specific Integrated Circuit) or a combination of hardware and software (e.g., a hardware/software co- design implementation) .
- the present invention can be used with a variety of image storage formats and image filtering techniques as well as a variety of feature detector and scorer techniques. Therefore, the pending claims are to encompass within their scope all such changes and modifications that fall within the true scope of the present invention.
Abstract
Description
Claims
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU21983/00A AU2198300A (en) | 1998-12-24 | 1999-12-20 | Improved techniques for spatial representation of data and browsing based on similarity |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US09/220,614 US6721759B1 (en) | 1998-12-24 | 1998-12-24 | Techniques for spatial representation of data and browsing based on similarity |
US09/220,614 | 1998-12-24 |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2000039712A1 WO2000039712A1 (en) | 2000-07-06 |
WO2000039712A9 true WO2000039712A9 (en) | 2001-07-12 |
Family
ID=22824244
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US1999/030298 WO2000039712A1 (en) | 1998-12-24 | 1999-12-20 | Improved techniques for spatial representation of data and browsing based on similarity |
Country Status (3)
Country | Link |
---|---|
US (2) | US6721759B1 (en) |
AU (1) | AU2198300A (en) |
WO (1) | WO2000039712A1 (en) |
Families Citing this family (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6578022B1 (en) * | 2000-04-18 | 2003-06-10 | Icplanet Corporation | Interactive intelligent searching with executable suggestions |
WO2001080065A2 (en) * | 2000-04-18 | 2001-10-25 | Icplanet Acquisition Corporation | Method, system, and computer program product for propagating remotely configurable posters of host site content |
US6681255B1 (en) | 2000-04-19 | 2004-01-20 | Icplanet Corporation | Regulating rates of requests by a spider engine to web sites by creating instances of a timing module |
AU2001253784A1 (en) * | 2000-04-25 | 2001-11-07 | Icplanet Acquisition Corporation | System and method for proximity searching position information using a proximity parameter |
WO2001082075A2 (en) * | 2000-04-25 | 2001-11-01 | Icplanet Acquisition Corporation | System and method for scheduling execution of cross-platform computer processes |
AU2001255610A1 (en) | 2000-04-25 | 2001-11-07 | Icplanet Acquisition Corporation | System and method related to generating and tracking an email campaign |
KR100824380B1 (en) * | 2002-08-08 | 2008-04-22 | 삼성전자주식회사 | Video recording/reproducing apparatus and method of displaying menu guide |
GB2395808A (en) * | 2002-11-27 | 2004-06-02 | Sony Uk Ltd | Information retrieval |
US8799883B2 (en) * | 2003-01-31 | 2014-08-05 | Hewlett-Packard Development Company, L. P. | System and method of measuring application resource usage |
US20060041375A1 (en) * | 2004-08-19 | 2006-02-23 | Geographic Data Technology, Inc. | Automated georeferencing of digitized map images |
JP2006107020A (en) * | 2004-10-04 | 2006-04-20 | Sony Corp | Content management system, content management method and computer program |
EP1839035A1 (en) * | 2005-01-13 | 2007-10-03 | E.I. Dupont De Nemours And Company | A method to specify acceptable surface appearance of a coated article |
US20060173874A1 (en) * | 2005-02-03 | 2006-08-03 | Yunqiang Chen | Method and system for interactive parameter optimization using multi-dimensional scaling |
US7392258B2 (en) * | 2005-02-25 | 2008-06-24 | International Business Machines Corporation | Method and computer program product for dynamic weighting of an ontological data model |
US7809754B2 (en) * | 2005-02-28 | 2010-10-05 | International Business Machines Corporation | Method and computer program product for generating a lightweight ontological data model |
US7707158B2 (en) * | 2005-02-28 | 2010-04-27 | International Business Machines Corporation | Method and computer program product for enabling dynamic and adaptive business processes through an ontological data model |
EP1755067A1 (en) * | 2005-08-15 | 2007-02-21 | Mitsubishi Electric Information Technology Centre Europe B.V. | Mutual-rank similarity-space for navigating, visualising and clustering in image databases |
US7295719B2 (en) * | 2005-08-26 | 2007-11-13 | United Space Alliance, Llc | Image and information management system |
EP1927058A4 (en) * | 2005-09-21 | 2011-02-02 | Icosystem Corp | System and method for aiding product design and quantifying acceptance |
JP4232774B2 (en) | 2005-11-02 | 2009-03-04 | ソニー株式会社 | Information processing apparatus and method, and program |
US8009910B2 (en) * | 2006-05-05 | 2011-08-30 | Valtion Teknillinen Tutkimuskeskus | Method, a system, a computer program product and a user interface for segmenting image sets |
US8244705B1 (en) * | 2008-02-22 | 2012-08-14 | Adobe Systems Incorporated | Rating system and spatial searching therein |
WO2010078925A1 (en) * | 2008-12-16 | 2010-07-15 | Attentio Sa/Nv | Method and system for monitoring online media and dynamically charting the results to facilitate human pattern detection |
US20100185672A1 (en) * | 2009-01-21 | 2010-07-22 | Rising Iii Hawley K | Techniques for spatial representation of data and browsing based on similarity |
US8750164B2 (en) * | 2010-07-06 | 2014-06-10 | Nicira, Inc. | Hierarchical managed switch architecture |
US9680750B2 (en) | 2010-07-06 | 2017-06-13 | Nicira, Inc. | Use of tunnels to hide network addresses |
US11636123B2 (en) * | 2018-10-05 | 2023-04-25 | Accenture Global Solutions Limited | Density-based computation for information discovery in knowledge graphs |
Family Cites Families (34)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4719571A (en) * | 1986-03-05 | 1988-01-12 | International Business Machines Corporation | Algorithm for constructing tree structured classifiers |
US5193185A (en) * | 1989-05-15 | 1993-03-09 | David Lanter | Method and means for lineage tracing of a spatial information processing and database system |
US5257365A (en) * | 1990-03-16 | 1993-10-26 | Powers Frederick A | Database system with multi-dimensional summary search tree nodes for reducing the necessity to access records |
CZ293613B6 (en) * | 1992-01-17 | 2004-06-16 | Westinghouse Electric Corporation | Method for monitoring the operation of a facility using CPU |
US5579471A (en) * | 1992-11-09 | 1996-11-26 | International Business Machines Corporation | Image query system and method |
US5574835A (en) * | 1993-04-06 | 1996-11-12 | Silicon Engines, Inc. | Bounding box and projections detection of hidden polygons in three-dimensional spatial databases |
JP2522154B2 (en) * | 1993-06-03 | 1996-08-07 | 日本電気株式会社 | Voice recognition system |
US5544352A (en) * | 1993-06-14 | 1996-08-06 | Libertech, Inc. | Method and apparatus for indexing, searching and displaying data |
US5710916A (en) * | 1994-05-24 | 1998-01-20 | Panasonic Technologies, Inc. | Method and apparatus for similarity matching of handwritten data objects |
AU4199396A (en) * | 1994-11-21 | 1996-06-17 | Oracle Corporation | Method and apparatus for multidimensional database using binary hyperspatial code |
US5664174A (en) * | 1995-05-09 | 1997-09-02 | International Business Machines Corporation | System and method for discovering similar time sequences in databases |
EP0788649B1 (en) * | 1995-08-28 | 2001-06-13 | Koninklijke Philips Electronics N.V. | Method and system for pattern recognition based on tree organised probability densities |
US5978794A (en) * | 1996-04-09 | 1999-11-02 | International Business Machines Corporation | Method and system for performing spatial similarity joins on high-dimensional points |
US5832182A (en) * | 1996-04-24 | 1998-11-03 | Wisconsin Alumni Research Foundation | Method and system for data clustering for very large databases |
US6047080A (en) * | 1996-06-19 | 2000-04-04 | Arch Development Corporation | Method and apparatus for three-dimensional reconstruction of coronary vessels from angiographic images |
IL119082A (en) * | 1996-08-16 | 2001-04-30 | Virtue Ltd | Method for creating graphic images |
US6085186A (en) * | 1996-09-20 | 2000-07-04 | Netbot, Inc. | Method and system using information written in a wrapper description language to execute query on a network |
JP3598183B2 (en) * | 1996-10-16 | 2004-12-08 | 株式会社東芝 | Multidimensional data management method, multidimensional data management device, medium recording multidimensional data management program |
DE29700549U1 (en) * | 1997-01-14 | 1997-05-07 | Trw Repa Gmbh | Adjustment device for a deflection fitting of a vehicle seat belt system |
US5963956A (en) * | 1997-02-27 | 1999-10-05 | Telcontar | System and method of optimizing database queries in two or more dimensions |
US6233575B1 (en) * | 1997-06-24 | 2001-05-15 | International Business Machines Corporation | Multilevel taxonomy based on features derived from training documents classification using fisher values as discrimination values |
US5974407A (en) * | 1997-09-29 | 1999-10-26 | Sacks; Jerome E. | Method and apparatus for implementing a hierarchical database management system (HDBMS) using a relational database management system (RDBMS) as the implementing apparatus |
US6070159A (en) * | 1997-12-05 | 2000-05-30 | Authentec, Inc. | Method and apparatus for expandable biometric searching |
US5987468A (en) * | 1997-12-12 | 1999-11-16 | Hitachi America Ltd. | Structure and method for efficient parallel high-dimensional similarity join |
US6148295A (en) * | 1997-12-30 | 2000-11-14 | International Business Machines Corporation | Method for computing near neighbors of a query point in a database |
JP3607107B2 (en) * | 1998-03-13 | 2005-01-05 | 株式会社東芝 | Data management device |
US6223182B1 (en) * | 1998-06-30 | 2001-04-24 | Oracle Corporation | Dynamic data organization |
US6567814B1 (en) * | 1998-08-26 | 2003-05-20 | Thinkanalytics Ltd | Method and apparatus for knowledge discovery in databases |
US6289354B1 (en) * | 1998-10-07 | 2001-09-11 | International Business Machines Corporation | System and method for similarity searching in high-dimensional data space |
US6301579B1 (en) * | 1998-10-20 | 2001-10-09 | Silicon Graphics, Inc. | Method, system, and computer program product for visualizing a data structure |
US6263334B1 (en) * | 1998-11-11 | 2001-07-17 | Microsoft Corporation | Density-based indexing method for efficient execution of high dimensional nearest-neighbor queries on large databases |
US6279007B1 (en) * | 1998-11-30 | 2001-08-21 | Microsoft Corporation | Architecture for managing query friendly hierarchical values |
US6492998B1 (en) * | 1998-12-05 | 2002-12-10 | Lg Electronics Inc. | Contents-based video story browsing system |
US6446068B1 (en) * | 1999-11-15 | 2002-09-03 | Chris Alan Kortge | System and method of finding near neighbors in large metric space databases |
-
1998
- 1998-12-24 US US09/220,614 patent/US6721759B1/en not_active Expired - Fee Related
-
1999
- 1999-12-20 AU AU21983/00A patent/AU2198300A/en not_active Abandoned
- 1999-12-20 WO PCT/US1999/030298 patent/WO2000039712A1/en active Application Filing
-
2003
- 2003-12-11 US US10/735,162 patent/US7496597B2/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
US7496597B2 (en) | 2009-02-24 |
US20040139096A1 (en) | 2004-07-15 |
AU2198300A (en) | 2000-07-31 |
WO2000039712A1 (en) | 2000-07-06 |
US6721759B1 (en) | 2004-04-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6721759B1 (en) | Techniques for spatial representation of data and browsing based on similarity | |
JP4516957B2 (en) | Method, system and data structure for searching for 3D objects | |
US7805440B2 (en) | System and method for simplifying and manipulating k-partite graphs | |
Pedersen et al. | Extending practical pre-aggregation in on-line analytical processing | |
De Marsicoi et al. | Indexing pictorial documents by their content: a survey of current techniques | |
US11687544B2 (en) | Adaptive analytics user interfaces | |
US20100185672A1 (en) | Techniques for spatial representation of data and browsing based on similarity | |
US11947596B2 (en) | Index machine | |
Chen et al. | Semantic models for multimedia database searching and browsing | |
KR20020006663A (en) | Multimedia archive description scheme | |
Arisawa et al. | Design of multimedia database and a query language for video image data | |
Berry et al. | AST: A library for modelling and manipulating coordinate systems | |
Tuijn et al. | CGOOD, a categorical graph-oriented object data model | |
CN112612933A (en) | Classified data visualization method | |
Hawkins et al. | Data structure fusion | |
Delest et al. | DAG-based visual interfaces for navigation in indexed video content | |
Noik | Challenges in graph-based relational data visualization. | |
KR100564739B1 (en) | The method for generating memory resident object-relational schema/query by using UML | |
CN116860901A (en) | Geographic knowledge association index and retrieval method based on graph network hierarchical division | |
Fang et al. | J3DV: A Java‐based 3D database visualization tool | |
Nakazato et al. | An interactive 3D visualization for content-based image retrieval | |
Belkhatir | An operational model based on knowledge representation for querying the image content with concepts and relations | |
Xie et al. | Operator-centric design patterns for information visualization software | |
Gardiner | A Framework for Processing viewer-based Directional Queries in a Simulated Mobile environment | |
Knoll et al. | A system for the content-based retrieval of textual and non-textual documents using a natural language interface |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG UZ VN YU ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH GM KE LS MW SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
AK | Designated states |
Kind code of ref document: C2 Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG UZ VN YU ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: C2 Designated state(s): GH GM KE LS MW SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG |
|
COP | Corrected version of pamphlet |
Free format text: PAGES 1-32, DESCRIPTION, REPLACED BY NEW PAGES 1-32; PAGES 33-41, CLAIMS, REPLACED BY NEW PAGES 33-41; DUE TO LATE TRANSMITTAL BY THE RECEIVING OFFICE |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
122 | Ep: pct application non-entry in european phase |