Modeling methods and systems
AssigneeUniversity of Basel
InventorsReinhard Tobias Knothe, Thomas Vetter
Life of this patent
Estimating the term…
Abstract
Methods and/or systems for modeling 3-dimensional objects (for example, human faces). In certain example embodiments, methods and/or systems usable for computer animation or static manipulation or modification of modeled images (e.g., faces), image processing, or for facial (or other object) recognition methods and/or systems.
Claims16
A method of representing, modeling, or coding object properties of a 3-dimensional object by using global and local components of a plurality of resolutions to adjust a 3-dimensional model of a 3-dimensional object comprising the following method steps:
- breaking said object properties into 3-dimensional modes based on spatial frequency;
- segmenting high frequency 3-dimensional modes into regional segments corresponding to specific features of said 3-dimensional object properties and performing a coarse-to-fine transition of the modes via an approximating subdivision scheme;
- computing a Principal Component Analysis (PCA) of said regional segments individually;
- combining a hierarchy of global and local components which are gradually decreasing in size of support into a linear statistical model;
- wherein, by performing said method, a global to local model of at least a portion of the 3-dimensional object is obtained, wherein said 3-dimensional model is computed using a 3-dimensional surface model of said 3-dimensional object from said single photograph using said global and local components;
- wherein global and local features are stacked into a single feature vector;
- wherein different masks are used for different levels.
A data processing system for representing, modeling, or coding object properties of a 3-dimensional object, comprising a display and an operating system, wherein one or more data processors or software implemented using one or more data processors are configured to adjust a 3-dimensional model of the 3-dimensional object using global and local components of a plurality of resolutions, wherein said adjusting comprises the following method steps:
- breaking said objects properties into 3-dimensional modes based on spatial frequency;
- segmenting high frequency 3-dimensional modes into regional segments corresponding to specific features of said 3-dimensional object properties and performing a coarse-to-fine transition of the modes via an approximating subdivision scheme;
- computing a Principal Component Analysis of said regional segments individually;
- combining a hierarchy of global and local components which are gradually decreasing in size of support into a linear statistical model, wherein the data processing system creates a global to local model of at least a portion of the 3-dimensional object, wherein said 3-dimensional model is computed using a 3-dimensional surface model of said 3-dimensional object from said single photograph using said global and local components;
- wherein global and local features are stacked into a single feature vector;
- wherein different masks are used for different levels.
Dependencies are inferred from the claim text.