Geometry-driven feature point-based image synthesis
First Claim
Patent Images
1. A computer implemented method for rendering a single frame of a synthesized image, comprising:
- generating a desired geometric component corresponding to a selected image for the single frame based on identified feature points from a set of representative images, where each image of the set has the identified feature points with a corresponding reference image geometric component, and wherein the geometric component is a dimensional vector of feature point positions; and
generating the selected image for the single frame from a composite of the set of representative images based on the desired geometric component;
wherein the selected image and each of the set of representative images comprises a plurality of subregions defined adjacent to each other in the single frame wherein adjacent subregions share a common boundary; and
wherein generating a desired geometric component is performed for each subregion, and wherein the composite of the set of representative images is based on the corresponding desired geometric component for each subregion, and the selected image includes a synthesized subregion for each subregion based on the composite by projecting the desired geometric component into the convex hull of the corresponding reference image geometric components to obtain a set of coefficient, where a coefficient is associated with a representative image, and wherein a texture of the corresponding representative image as a function of each of the associated coefficients and blending at least some boundaries between adjacent subregions of the selected image order to generate the selected image.
2 Assignments
0 Petitions
Accused Products
Abstract
A method and system uses geometry-driven feature point analysis to synthesize images including for example facial expressions. Given the feature point positions (geometry) of an expression, the method automatically synthesizes the corresponding expression image, which has photorealistic and natural looking expression details.
27 Citations
30 Claims
-
1. A computer implemented method for rendering a single frame of a synthesized image, comprising:
-
generating a desired geometric component corresponding to a selected image for the single frame based on identified feature points from a set of representative images, where each image of the set has the identified feature points with a corresponding reference image geometric component, and wherein the geometric component is a dimensional vector of feature point positions; and generating the selected image for the single frame from a composite of the set of representative images based on the desired geometric component;
wherein the selected image and each of the set of representative images comprises a plurality of subregions defined adjacent to each other in the single frame wherein adjacent subregions share a common boundary; and
wherein generating a desired geometric component is performed for each subregion, and wherein the composite of the set of representative images is based on the corresponding desired geometric component for each subregion, and the selected image includes a synthesized subregion for each subregion based on the composite by projecting the desired geometric component into the convex hull of the corresponding reference image geometric components to obtain a set of coefficient, where a coefficient is associated with a representative image, and wherein a texture of the corresponding representative image as a function of each of the associated coefficients and blending at least some boundaries between adjacent subregions of the selected image order to generate the selected image. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 27, 28, 29)
-
-
17. A computer implemented method for rendering a single frame of a synthesized image based on feature points, comprising:
-
accessing a set of stored representatives of various images, wherein each image of the set of the images has the same corresponding feature points associated therewith and a corresponding reference image geometric component and wherein the geometric component is a dimensional vector of feature point positions; ascertaining a position of at least one feature point from a change in position of another feature point based on a change in movement of the selected feature point and based on the set of stored representatives of various images; and rendering a new image for the single frame with two or more feature points having changed position;
wherein the new image and each image in the set of stored representatives of various images comprises a plurality of subregions defined adjacent to each other wherein each subregion of the new image has associated therewith a desired geometric component, wherein adjacent subregions share a common boundary, and rendering the new image includes rendering a synthesized subregion for each subregion by projecting the desired geometric component into the convex hull of the corresponding reference image geometric components to obtain a set of coefficients, where a coefficient is associated with a representative image, and wherein a texture of the subregion of the new image is based on combining the corresponding texture of the corresponding representative image as a function of each of the associated coefficients and blending at least some boundaries between adjacent subregions in the new image. - View Dependent Claims (18, 19, 20, 21, 22, 23, 30)
-
-
24. A computer implemented method for rendering a single frame of a synthesized image based on feature points, comprising:
-
rendering a image with identified feature points, wherein the feature points are grouped in hierarchical sets, each set pertaining to a different region of the image; receiving information indicative of a user moving a selected feature point; accessing a set of stored representatives of various images, wherein each image of the set of stored images has the same corresponding feature points associated therewith and a corresponding reference image geometric component, and wherein the geometric component is a dimensional vector of feature point positions; ascertaining a position of at least one feature point in each of two different sets from a change in position of another feature point based on a change in movement of the selected feature point and based on the set of stored representatives of various images, wherein ascertaining comprises ascertaining positions of feature points in a set from a change in position of one feature point int he set using principle components derived from analyzing positional changes of feature points in the set of stored representatives of various images through principle component analysis; and rendering a new image for the single frame with two or more feature points having changed position in two sets;
wherein the new image and each image in the set of stored representatives of various images comprises a plurality of subregions adjacent to each other wherein each subregion of the new image has associated therewith a desired geometric component, wherein adjacent subregions share a common boundary, and rendering a synthesized subregion for each subregion in the new image is obtained by protecting t he desired geometric component into the convex hull of t he corresponding reference image geometric components to obtain a set of coefficients, where a coefficient is associated with a representative image, and wherein a texture of the selected image is based on combining the corresponding texture of the corresponding representative image as a function of each of the associated coefficients. - View Dependent Claims (25, 26)
-
Specification