Adaptive modeling and segmentation of visual image streams
First Claim
1. A method for developing a scene model from a visual image sequence containing a sequence of visual image frames, the visual image sequence including a visual representation of one or more visual objects, the method comprising the steps of:
- (a) analyzing portions of the visual image sequence in accordance with input parameters using a machine vision process by performing the steps of;
(i) defining an image-based data object in an image-based model of the visual image sequence containing a pixel representation corresponding to a portion of at least one frame of the visual image sequence;
(ii) defining an abstraction-based data object in an abstraction-based model of the visual objects containing an abstract representation of at least a portion of one of the visual objects represented by the visual image sequence; and
(iii) storing a link data object in a correlation mesh data structure indicating a correspondence between the image-based data object and the abstraction-based data object;
(b) refining an abstraction-based model of the visual imago sequence by performing the steps of;
(i) accepting input parameters to define in the abstraction-based model of the visual objects a new abstraction-based data object containing a new abstract representation of at least a portion of one of the visual objects contained in the visual image sequence that differs from the abstract representations contained in abstraction-based objects defined in the analysis step (a); and
(ii) adding a link object in the correlation mesh data structure indicating a correspondence between the new abstraction-based data object and another data object defined in the scene model; and
(c) iteratively improving the scene model by performing certain selected ones of steps (a) through (b) in an order as specified by user input until a desired level of refinement is obtained in the scene model such that selected link objects in the correlation mesh data structure added in iterations of the refining step (b) are used to provide additional input parameters to subsequent iterations thereby allowing the scene model to converge.
2 Assignments
0 Petitions
Accused Products
Abstract
A technique for converging upon a computer-based model of a real world or synthetic scene. The computer model makes use of abstraction-based data objects as well as image-based data objects. A correlation mesh provides links between related image-based and abstraction-based objects. An initial step in a process analyzes an input image stream and user inputs to derive initial image-based objects and abstraction-based objects for the scene model. Subsequent steps in the process allow user inputs to refine the image-based objects, abstraction-based objects and/or the correlation mesh. As a result, refinements to the image-based object model of the scene can improve the abstraction-based model of the scene, and refinements to the abstraction-based object model can improve the image-based model.
250 Citations
43 Claims
-
1. A method for developing a scene model from a visual image sequence containing a sequence of visual image frames, the visual image sequence including a visual representation of one or more visual objects, the method comprising the steps of:
-
(a) analyzing portions of the visual image sequence in accordance with input parameters using a machine vision process by performing the steps of; (i) defining an image-based data object in an image-based model of the visual image sequence containing a pixel representation corresponding to a portion of at least one frame of the visual image sequence; (ii) defining an abstraction-based data object in an abstraction-based model of the visual objects containing an abstract representation of at least a portion of one of the visual objects represented by the visual image sequence; and (iii) storing a link data object in a correlation mesh data structure indicating a correspondence between the image-based data object and the abstraction-based data object; (b) refining an abstraction-based model of the visual imago sequence by performing the steps of; (i) accepting input parameters to define in the abstraction-based model of the visual objects a new abstraction-based data object containing a new abstract representation of at least a portion of one of the visual objects contained in the visual image sequence that differs from the abstract representations contained in abstraction-based objects defined in the analysis step (a); and (ii) adding a link object in the correlation mesh data structure indicating a correspondence between the new abstraction-based data object and another data object defined in the scene model; and (c) iteratively improving the scene model by performing certain selected ones of steps (a) through (b) in an order as specified by user input until a desired level of refinement is obtained in the scene model such that selected link objects in the correlation mesh data structure added in iterations of the refining step (b) are used to provide additional input parameters to subsequent iterations thereby allowing the scene model to converge. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 41, 42, 43)
-
-
32. A method for developing a computerized scene model from a digital image, the method comprising the steps of:
-
(a) running an automated image analysis algorithm according to input parameters to provide image analysis results data; (b) storing portion of the image analysis results data as an image-based data object in the scene model; (c) storing other portions of the image analysis results data as an abstraction-based data object in the scene model; (d) annotating a correlation mesh data object in the scene model with data indicating a link between the image-based data object and the abstraction-based data object; (e) refining the scene model by accepting input parameters concerning at least one new abstract representation of the digital image, the new abstract object containing an abstract representation differing from the abstraction-based object stored in step (c); (f) annotating a correlation mesh data object in the scene model with data indicating a link between the new abstraction-based data object and other data objects in the scene model; and (g) iterating selected ones of steps (a) through (f) in response to user input until a desired level of refinement is obtained in the scene model such that selected link objects in the correlation mesh data structure added in iterations of the refining step (e) are used to provide additional input parameters to subsequent iterations, thereby allowing the scene model to converge. - View Dependent Claims (33, 34, 35, 36, 37, 38, 39, 40)
-
Specification