JP2005523488A - Automatic 3D modeling system and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a 3D modeling system and method, and more particularly, a system and method for combining image-based automatic model generation technology with interactive and real-time character orientation technology to provide rapid creation of virtual 3D humans. To do.
An automated 3D modeling system and method capable of generating a 3D model from a photograph or other image. For example, a 3D model of a human face can be automatically generated. The system and method also allows automatically generating gestures / behaviors associated with 3D models so that gestures / behaviors can be applied to any 3D model.

Description

RELATED APPLICATIONS This application is a priority from US Provisional Application Serial No. 60 / 312,384 entitled "Automatic 3D Modeling System and Method" filed 14 August 2001, which is incorporated herein by reference. We claim rights under "35 USC § 119".
The present invention relates to 3D modeling systems and methods, and more particularly, systems and methods that combine image-based automatic model generation techniques with interactive and real-time character orientation techniques to provide rapid creation of virtual 3D humans. About.

  There are many different techniques for generating a moving image of a three-dimensional object on a computer display. At first, the video was not very good quality, so the video screen (eg, face) looked like a character made of wood. In particular, the user will generally see the face of a movie, but its features and expressions are static. Perhaps the mouth opens and closes and the eyes may blink, but the facial expressions and videos in general resemble a wooden doll. The problem is that these videos were generally created as drawings from scratch and were not rendered using basic 3D models to capture a more realistic look, so the videos were not realistic and not much It did not appear to be alive. More recently, animation has been improved so that the skin of a person can be covered with skin to provide a more realistic animation screen.

  To capture a more realistic look to the video, such video is now rendered on one or more deformation grids, but the video is still rendered by a professional company and redistributed to the user Often done. This results in a high quality video, but is limited in that the user does not have the ability to customize a specific video, for example the user's own video, for use as a virtual person. As the functionality of the “Internet” or “World Wide Web” advances, these virtual people will expand the capabilities and interaction between users. Accordingly, it would be desirable to provide a 3D modeling system and method that allows a typical user to quickly and easily generate a 3D model from an image useful as a virtual person such as a photograph.

US Patent Provisional Application Serial No. 60 / 312,384

  With a skilled video producer creating a model, a typical system also requires that same video producer be required to animate various gestures that they typically want to give to that model. did. For example, a video creator would create a video that smiles, shakes, or speaks, and then incorporates these videos into the model to give the model a target gesture. The process of generating behavior / gesture data is slow and expensive and requires a skilled video producer. It would be desirable to provide an automatic mechanism for generating gestures and behavior for a model without the assistance of a skilled animation producer. It is these end goals that the present invention aims for.

  The present invention generally provides an image processing technique that allows a photo-like realistic 3D model of an object such as a human face to be automatically generated from a single image (or multiple images), Use statistical analysis and 3D geometric deformation. For example, for a human face, facial features and feature details from a single photo (or series of photos) are identified and used to generate an appropriate 3D model. Image processing techniques and texture mapping techniques also optimize the method of using one or more photographs as detailed and photo-realistic textures for 3D models.

  According to another aspect of the invention, a human gesture can be captured and abstracted so that it can be applied to any other model. For example, a lively smile of a specific person can be captured. The smile can then be transformed into a feature space that provides gesture abstraction. An abstracted gesture (eg, movement of different parts of the model) is captured as one gesture. This gesture can then be used for any other model. That is, according to the present invention, the system enables the generation of gesture models that can be used with other models.

  According to the present invention, a method for generating a three-dimensional model of an object from an image is provided. The method includes determining the boundaries of the object to be modeled and determining the location of one or more landmarks on the object to be modeled. The method further includes determining a scale and orientation of the object in the image based on the position of the landmark, aligning the image of the object having the landmark with the deformation grid, and an object to the deformation grid. Generating a 3D model of the object based on the mapping of the image of the object.

  According to another aspect of the invention, a computer-implemented system for generating a three-dimensional model of an image is provided. The system includes a 3D model generation module further comprising instructions for receiving an image of the object and instructions for automatically generating a 3D model of the object. The system further includes instructions for generating a feature space and instructions for generating a gesture object corresponding to the gesture of the object so that the gesture behavior can be applied to another model of the object. A gesture generating module.

  According to yet another aspect of the invention, a method for automatically generating an automatic gesture model is provided. The method includes receiving an image of an object to perform a specific gesture, and determining a movement associated with the gesture from the movement of the object to generate a gesture object, the gesture object further comprising: A color change variable that memorizes the color change that occurs during the gesture, a two-dimensional change variable that memorizes the surface change that occurs during the gesture, and a three-dimensional change that memorizes the vertex change associated with the object during the gesture Including variables.

  According to yet another aspect of the present invention, a gesture object data structure is provided for storing data associated with a gesture related to an object. Gesture objects store texture change variables that store model color changes between gestures, texture map change variables that store model surface changes between gestures, and model vertex changes between gestures Vertex change variables, texture change variables, texture map change variables, and vertex change variables allow gestures to be applied to another model having textures and vertices. The gesture object data structure stores its data in a vector space where many individual instances of the model can use color schemes, surface motions, and 3D motions.

  The present invention has a wide range of uses, which will be described below in the context of generating a 3D model of a person's face and gestures that accompanies the human reality. Those skilled in the art will recognize that any other 3D models and gestures can be generated using the principles and techniques described herein, and thus the following description is simply the present invention. The specific application is merely illustrative, and the present invention is not limited to the facial model described herein.

  In order to generate a 3D model of a human face, the present invention preferably employs a series of complex image processing techniques to determine a set of landmark points 10 that serve as a guide for generating a 3D model. Execute. FIG. 1 is a flow diagram illustrating a preferred algorithm for generating a 3D model of a human face. Referring to FIG. 1, an image acquisition process (stage 1) is used to read one or more photographs (or other images) of a person's face (eg, “face photograph”) into computer memory. . The image can preferably be read as a “JPEG” image, but other image type formats can also be used without departing from the invention. Images are read from diskettes, downloaded from the “Internet”, or otherwise known techniques so that the image processing techniques of the present invention can be performed on the images to generate a 3D model Can be read into memory.

Since different images may have different orientations, the proper orientation of the image should be determined by finding and grading the appropriate landmark point 10. Determining the orientation of the image allows for a more realistic rendering of the image on the deformation grid. Here, the positioning of the appropriate mark point 10 will be described in detail below.
Referring to FIG. 1, to find the landmark point 10 on the image, preferably to remove the variable background of the image so that the head boundary (in the case of a face) can be separated on the image. Can perform a “seed fill” operation on the image (step 2). FIG. 4 is a typical image 20 of a person's head that can be read into computer memory during the image acquisition process (FIG. 1, stage 1). The “seed fill” operation (FIG. 1, stage 2) is a known iterative paint fill operation, which is based on, for example, the background 24 of the image 20 based on the color and brightness of one or more points 22. This is accomplished by identifying one or more points 22 within and extending a paint fill area 26 of similar color and brightness outward from point 22. Preferably, the “seed-fill” operation better replaces the image color and lightness background 24 with an opaque background so that the head boundary can be more easily determined.

  Referring again to FIG. 1, the boundary of the head 30 finds, for example, the vertical center of the image (line 32) and integrates over the horizontal area 34 from the center line 32 to determine the width of the head 30 (nonfill). Find the horizontal center (line 36) of the image and integrate over the vertical area 38 from the center line 36 to determine the height of the head 30 (using a non-fill operation). (Step 3). In other words, a statistically directed linear integration of pixel fields with different values based on the presence of an object or the presence of a background is performed. This is illustrated in FIG. 5 which shows the exemplary image 20 of FIG. 4 with an opaque background 24.

  Referring again to FIG. 1, once the width and height of the head 30 are determined, the boundary of the head 30 is determined by the statistical characteristics of the height of the head 30, the integrated horizontal area 34 and the head 30. This can be determined by using the known properties with the top. Typically, the height of the head is about 2/3 of the image height, and the width of the head will be about 1/3 of the image width. The height of the head may also be 1.5 times the width of the head, which is used as a first approximation.

  With the boundary of the head 30 determined, the position of both eyes 40 can be determined (step 4). Since both eyes 40 are typically located in the upper half of the head 30, statistical calculations can be used, and to separate the boundary areas 46 a and 46 b of the eyes, the head boundary is separated from the upper half 42. It can be divided into a lower half 44. The upper half 42 of the head boundary can be further divided into a right part 46a and a left part 46b, respectively, to separate the left eye 40a and the right eye 40b. This is illustrated in detail in FIG. 6, which shows the exemplary image 20 of FIG.

Still referring to FIG. 1, the most central region of each eye 40a and 40b can be found by identifying the high contrast luminance circular region 48 within the respective eye boundary 46a and 46b (step 5). . This operation can be performed repeatedly from the central point 48 across the bounded areas 46a and 46b, and the results are graded to determine the proper boundaries of both eyes 40a and 40b. can do. FIG. 7 shows the exemplary image of FIG. 6 with the high-contrast luminance portion of both eyes identified by a dashed line.
Referring again to FIG. 1, with both eyes 40a and 40b identified, the line 50 connecting both eyes 40a and 40b is analyzed and the offset angle of the line 50 from the horizontal axis of the screen is determined to determine the head angle. The scale and orientation of the unit 30 can be determined (step 6). The scale of the head 30 can be derived from the boundary width according to the equation boundary width / model width.

  When the above information has been determined, the approximate landmark point 10 on the head 30 can be appropriately identified. Preferred mark points 10 are: a) outer head boundaries 60a, 60b and 60c, b) inner head boundaries 62a, 62b, 62c and 62d, c) right eye boundaries 64a to 64d and left eye boundaries 64w to 64z, d. Those skilled in the art will appreciate that)) nose corners 66a and 66b, and e) mouth corners 68a and 68e (mouth lines), although other landmarks may be used without departing from the invention. Will admit. FIG. 8 exemplarily shows the above-described mark points shown for the image of FIG.

  Once the appropriate landmark position 10 on the head 30 has been determined, the image can be properly aligned with one or more deformation grids (described below) that form the 3D model 70 of the head (see below). Step 7). In the following, some deformation grids that can be used to form the 3D model 70 are described, but these are just some examples of deformation grids that can be used to form the 3D model. However, those skilled in the art will recognize that other deformed grids may be used without departing from the present invention. FIG. 9 shows an example of a 3D model of a human face generated using the 3D model generation method according to the present invention. Here, the 3D model generation system will be described in more detail below.

  FIG. 2 shows an example of a computer system 70 that can implement the 3D model generation method and the gesture model generation method. In particular, the 3D model generation method and the gesture model generation method can be implemented as one or more portions of software code (or compiled software code) executed by a computer system. The method according to the invention can also be implemented on a hardware device in which the method according to the invention is programmed in a hardware device. Returning to FIG. 2, the computer system 70 shown in the figure is a personal computer system. However, the present invention can be implemented on a variety of different computer systems, such as client / server systems, server systems, workstations, etc., and is not limited to implementation on any particular computer system. . The illustrated computer system includes a display device 72, such as a CRT or LCD, a chassis 74, and one or more inputs such as the illustrated keyboard 76 and mouse 78 that allow the user to interact with the computer system. An output device can be included. For example, a user can input data or commands to a computer system using a keyboard or mouse, and output data from the computer system using a display device (visible data) or a printer (not shown). Can be received. The chassis 74 can contain the computing resources of the computer system and, as is known, one or more central processing units (CPUs) 80 that control the operation of the computer system and is not powered by the computer system. Persistent storage devices 82, such as hard disk drives, optical disk drives, tape drives, etc., that store data and instructions executed by the “CPU” at times, and data currently being executed by the “CPU”, as is known And a memory 84 such as a “DRAM” that temporarily stores instructions and loses the data when the computer system is not powered. In order to implement the 3D model generation and gesture model generation method according to the present invention, the memory is a 3D modeler that is a series of instructions and data executed by the “CPU” 80 to execute the 3D model and gesture generation method described above. 86 can be stored. Here, the 3D modeler will be described in more detail below.

  FIG. 3 is a more detailed view of the 3D modeler 86 shown in FIG. In particular, the 3D modeler includes a 3D model generation module 88 and a gesture generation module 90, each implemented using one or more computer program instructions. 12A-12B and FIGS. 14A-14B show pseudocode that can be used to implement each of these modules. As shown in FIG. 3, an image of an object such as a human face is input to a system as shown in the figure. The images are supplied to the 3D model generation module and the gesture generation module as shown. The output from the 3D model generation module is a 3D model of the image automatically generated as described above. The output from the gesture generation module is one or more gesture models that are then applied to and used on any 3D model, including any model generated by the 3D model generation module. Can do. The gesture generator will be described in detail later with reference to FIG. In this way, the system allows a 3D model of any object to be quickly generated and implemented. In addition, the gesture generator allows one or more gesture models, such as a smiling gesture or a waving gesture, to be automatically generated from a particular image. The advantage of a gesture generator is that this gesture model can then be applied to any 3D model. Gesture generators also eliminate the need for skilled video creators to perform gestures. Here, the deformation grid for generating the 3D model will be described below.

  10A-10D show an exemplary deformed grid that can be used to form a 3D model 70 of a human head. FIG. 10A shows a boundary space deformation grid 72, which is preferably the innermost deformation grid. A feature space deformation grid 74 (shown in FIG. 10B) is overlaid on the boundary space deformation grid 72. The edge space deformation grid 76 (shown in FIG. 10C) preferably overlies the feature space deformation grid 74. FIG. 10D shows a detailed deformation grid 78, which is preferably the outermost deformation grid.

  The grid preferably has a landmark position 10 (see FIG. 10E) so that the head image will be substantially aligned with the deformation grid when the landmark position 10 of the head image 30 is aligned with the landmark position 10 of the deformation grid. Aligned). In order to properly align the head image 30 with the deformation grid, the user can “drag” specific landmarks to different areas on the image 30 using, for example, a mouse or other input device. The mark position accuracy on the image 30 can be manually refined (step 8). Using the new landmark position information, if appropriate, the image 30 can be modified with respect to the deformation grid in order to properly align the head image 30 with the deformation grid (step 9). The state of the new model can then be calculated, the detail grid 78 can then be removed (stage 10), the behavior can be scaled against the resulting 3D model (stage 11), and the virtual The model can be stored for use as a person (step 12). Here, the automatic gesture generation according to the present invention will be described in more detail below.

  FIG. 11 is a flowchart illustrating an automatic gesture model generation method 100 according to the present invention. In general, automatic gesture generation results in a gesture object that can then be applied to any 3D model so that gesture behavior can be quickly generated and reused in other models. Can do. Usually, separate gesture models will be required for different types of 3D models. For example, smile gestures may need to be automatically generated for men, women, boys, and girls in order to make the gestures more realistic. The method begins at step 102 where a common feature space is generated. The feature space is a common space used to store and represent an object scaler that captures images of objects such as faces, object movement during gestures, and differences between different objects. Gesture objects generated using the method also store a scaler field variable that stores a mapping between model space and feature space that allows transformation of motion and geometric data. The automatic gesture generation method uses a specific image of the object such as the face to generate an abstraction of the gesture of the object such as a smile, which is then stored as a gesture object, which then It involves steps that allow it to be applied to any 3D model.

  Returning to FIG. 11, the method determines a correlation between the feature space and the image space at step 104 to determine a texture map change that represents a change to the surface motion of the image during the gesture. . In step 106, the method updates the texture map from the image (to check correlation), applies the resulting texture map to the feature space, and stores the texture map changes, as shown in FIGS. 14A and 14B. A variable “stDeltaChange” as shown in a typical pseudo code is generated. In step 108, the method determines a change in the 3D vertex of the image model during the gesture that captures the 3D movement that occurs during the gesture. In step 110, the vertex changes are applied to the feature space and captured in a variable “VertDeltaChange” in the gesture object, as shown in FIGS. 14A and 14B. In step 112, the method determines the texture color scheme that occurs during the gesture and applies it to the feature space. The texture color scheme is captured in a “DeltaMap” variable in the gesture object. In step 114, a gesture object is created that includes the “stDeltaChange”, “VertDeltaChange”, and “DeltaMap” variables that include the color scheme that occurs during the gesture and the 2D and 3D movements. These variables represent only the movements and color changes that occur during the gesture so that the gesture object can then be applied to any 3D model. In short, a gesture object can extract a gesture that exists in a particular image model and turn it into an abstract object that contains the essential elements of that gesture, so that it can then be applied to any 3D model. It can be so.

The gesture object also includes a scaler field variable that stores a mapping between the feature space of the gesture and the model space of the model to allow conversion of geometric and motion data. The “scalerArray” has an entry for each geometric vertex in the “gesture” object. Each entry is a three-dimensional vector that holds the change in scale from its undeformed state to its deformed state for that vertex at the “feature” level. The scale is calculated by evaluating the change in scaler by the vertices in the “feature” space at the distance from that vertex to each connected vertex. The scaler for a given “gesture” vertex is calculated by weighted interpolation of its “vertex” position when mapped to a polygonal “UV” space at “feature level”. The shape and size of the polygon at the feature level is selected to match the similarly scaled area of motion. This is determined by analyzing the visual flow of typical facial gestures. The above method is illustrated in more detail in the pseudo code shown in FIGS. 14A and 14B.
FIGS. 12A-12B and FIGS. 13A-13B each include a sample pseudocode algorithm and an exemplary workflow process for automatically generating a 3D model in accordance with the present invention.

  The automatically generated model can incorporate built-in behavior animation and interactivity. For example, for a human face, such expressions include gestures, mouth positions for lips synchronization, and head movements. Such behavior can be integrated using techniques such as automatic lip synchronization, speech synthesis, natural language processing, and speech recognition to initiate or be initiated by a user or data driven event. Is possible. For example, an automatically generated model real-time lip synchronization can be associated with an audio track. In addition, real-time analysis of speech spoken by intelligent agents can be provided, and synchronized head and face gestures are initiated to provide automatic and lively movements as speech is delivered.

  That is, virtual humans are used as interactive responsive front ends for information contained in knowledge bases, customer resource management systems, learning management systems, and entertainment applications and communications through chat, instant messaging, and email Can be arranged to work as an intelligent agent that can. Here, an example of a gesture generated from an image of a 3D model and then applied to another model according to the present invention will be described below.

  FIG. 15 shows an example of a basic 3D model for kristen, which is the first model. The 3D model shown in FIG. 15 was previously generated as described above using the 3D model generation process. FIG. 16 shows the second 3D model generated as described above. These two models describe automatically generating a smiling gesture from an existing model to generate a gesture object, and then applying the gesture object thus generated to another 3D model Will be used to do. FIG. 17 shows an example of the first model in the neutral gesture, and FIG. 18 shows an example of the first model in the smile gesture. Next, the smile gesture of the first model is captured as described above. FIG. 19 shows an example of a smile gesture map (graphic version of the gesture object described above) generated from the first model based on the neutral gesture and the smile gesture. As described above, the gesture map abstracts the gesture behavior of the first model into a series of color scheme changes, texture map changes, and 3D vertex changes, which are then any other having texture maps and 3D vertices. It can be applied to 3D models. This gesture map (including the variables described above) can then be used to apply the gesture object to another model according to the present invention. In this way, the automatic gesture generation process allows various gestures for a 3D model to be abstracted and then applied to other 3D models.

  FIG. 20 is an example of a feature space in which both models are superimposed on each other to show that the feature spaces of the first and second models are aligned with each other. Here, the application of the gesture map (and hence the gesture object) to another model will be described in more detail below. In particular, FIG. 21 shows a neutral gesture of the second model. FIG. 22 shows a smile gesture (a gesture generated by the first model) applied to the second model to give a smile gesture to the second model even when the second model is not actually showing a smile. From the map).

  As described above, a particular method for finding a point of a landmark location on an image and a particular method for generating a gesture have been described, but other methods may be used without departing from the invention as defined by the claims. Those skilled in the art will recognize that techniques may be used. For example, a technique such as pyramid transformation that uses frequency analysis of the image by down-sampling each level and analyzing the frequency difference at each level can be used. In addition, other techniques such as side sampling and image pyramid techniques can be used to process the image. Furthermore, orthogonal (low-pass) filtering techniques can be used to increase the signal strength of facial features, and fuzzy logic techniques can be used to determine the overall position of the face. The location of the landmark can then be determined by a known corner finding algorithm.

3 is a flowchart illustrating a method for generating a 3D model of a human face. FIG. 2 illustrates an example of a computer system that can be used to implement a 3D modeling method according to the present invention. 1 is a block diagram illustrating a 3D model generation system according to the present invention in more detail. FIG. FIG. 3 is a diagram illustrating a typical image of a person's head that can be loaded into a computer's memory during an image acquisition process. FIG. 5 illustrates the exemplary image of FIG. 4 with an opaque background after processing the image with a “seed fill” operation. FIG. 6 shows the exemplary image of FIG. 5 with a dashed line indicating a particular boundary area around the position of both eyes. FIG. 7 shows the exemplary image of FIG. 6 with high contrast brightness portions of both eyes identified by dashed lines. FIG. 6 is an exemplary diagram illustrating points at various landmark positions relative to a human head. It is a figure which shows an example of the 3D model of the human face by this invention. FIG. 3 shows one of each deformation grid that can be used to generate a 3D model of a human head. FIG. 3 shows one of each deformation grid that can be used to generate a 3D model of a human head. FIG. 3 shows one of each deformation grid that can be used to generate a 3D model of a human head. FIG. 3 shows one of each deformation grid that can be used to generate a 3D model of a human head. It is a figure which shows the deformation | transformation grid overlaid on each other. 4 is a flowchart illustrating an automatic gesture behavior generation method according to the present invention. FIG. 3 is a diagram showing typical pseudo code for executing the image processing technique of the present invention. FIG. 3 is a diagram showing typical pseudo code for executing the image processing technique of the present invention. FIG. 6 is a diagram illustrating an exemplary operation flow process for automatically generating a 3D model according to the present invention. FIG. 6 is a diagram illustrating an exemplary operation flow process for automatically generating a 3D model according to the present invention. FIG. 3 illustrates exemplary pseudo code for executing an automatic gesture behavior model in accordance with the present invention. FIG. 3 illustrates exemplary pseudo code for executing an automatic gesture behavior model in accordance with the present invention. It is a figure which shows the example of the basic 3D model with respect to kristen which is a 1st model. It is a figure which shows the example of the basic 3D model with respect to a 2nd model and Ellie. It is a figure which shows the example of the 1st model which made the neutral gesture. It is a figure which shows the example of the 1st model which made the gesture of smile. It is a figure which shows the example of the smile gesture map produced | generated from the neutral gesture and smile gesture of the 1st model. It is a figure which shows the example of the feature space with which both models were piled up on each other. It is a figure which shows the example of the neutral gesture with respect to a 2nd model. FIG. 6 is a diagram illustrating an example of a smile gesture generated from a first model applied to a second model to generate a smile gesture on the second model.

Explanation of symbols

86 3D Modeler 88 3D Model Generation Module 90 Gesture Generation Module

Claims (13)

A method for generating a three-dimensional model of an object from an image,
Determining the boundaries of the object being modeled;
Determining the position of one or more landmarks on the object to be modeled;
Determining the scale and orientation of the object in the image based on the position of the landmark;
Aligning an image of the object having the landmark with a deformation grid;
Generating a 3D model of the object based on a mapping of the image of the object to the deformation grid;
A method comprising the steps of:   The method of claim 1, wherein the step of determining the boundary further comprises a statistically directed linear integration of a field of pixels having different values depending on the presence of an object or the presence of a background.   The method of claim 1, wherein determining the boundary further comprises performing a statistically oriented seed fill operation to remove background around an image of the object.   Determining the landmark includes identifying features found by procedural correlation or bandpass filtering, and thresholding within a statistically characterized area as determined during the boundary determination. The method of claim 3, further comprising the step of:   The method of claim 4, wherein determining the landmark further comprises determining an additional landmark based on refinement of the boundary area.   The method of claim 5, wherein determining the landmark further comprises adjusting the landmark by a user. A computer-implemented system for generating a three-dimensional model of an image,
A 3D model generation module further comprising: an instruction for receiving an image of the object; and an instruction for automatically generating a 3D model of the object;
Instructions for generating a feature space, and instructions for generating a gesture object corresponding to the gesture of the object so that the behavior of the gesture can be applied to a model of another object; A gesture generation module further comprising:
A system characterized by including. A method for automatically generating an automatic gesture model,
Receiving an image of an object performing a specific gesture;
Determining a movement associated with the gesture from the movement of the object to generate a gesture object;
Including
The gesture object includes a color change variable that stores a color change that occurs during the gesture, a two-dimensional change variable that stores a surface change that occurs during the gesture, and the object between the gestures. Further including a three-dimensional change variable for storing vertex changes associated with
A method characterized by that.   The method of claim 8, further comprising generating a feature space to which the gesture is mapped during the automatic gesture generation process.   The method of claim 9, wherein determining the movement further comprises determining a correlation between the feature space and an image of the object.   The method of claim 9, further comprising transforming the geometric and motion vectors into and out of the feature space.   10. The method of claim 9, further comprising applying color scheme, texture motion, and geometric motion changes from one model to another using the feature space. A gesture object data structure for storing data associated with a gesture of an object,
A scalar field variable that stores a mapping between the feature space of the gesture and the model space of the model to enable transformation of geometric and motion data;
Texture change variables that memorize changes in the model's color scheme between gestures;
A texture map change variable that stores changes in the surface of the model during the gesture;
A vertex change variable that stores a change in the model vertex during the gesture;
Including
The texture change variable, the texture map change variable, and the vertex change variable allow the gesture to be applied to another model having a texture and a vertex.
A structure characterized by that.

Images