Optimal gradient pursuit for image alignment
AssigneeGeneral Electric Company
InventorsXiaoming Liu, Frederick Wilson Wheeler, Peter Henry Tu, Jilin Tu
Life of this patent
Estimating the term…
Abstract
A method for image alignment is disclosed. In one embodiment, the method includes acquiring a facial image of a person and using a discriminative face alignment model to fit a generic facial mesh to the facial image to facilitate locating of facial features. The discriminative face alignment model may include a generative shape model component and a discriminative appearance model component. Further, the discriminative appearance model component may have been trained to estimate a score function that minimizes the angle between a gradient direction and a vector pointing toward a ground-truth shape parameter. Additional methods, systems, and articles of manufacture are also disclosed.
Claims20
A method comprising:
- acquiring a facial image of a person;
- and
- via software executed by a processor of a system, using a discriminative face alignment model to align a generic facial mesh to the facial image to facilitate locating of facial features of the facial image, the discriminative face alignment model including a generative shape model component and a discriminative appearance model component, the discriminative appearance model component having been trained with training data to estimate a score function that is a function of a shape parameter of a given image and that seeks to minimize an angle between a gradient direction of the score function for the shape parameter and an ideal alignment travel direction for the shape parameter.
A system comprising:
- a memory device having a plurality of routines stored therein;
- a processor configured to execute the plurality of routines stored in the memory device, the plurality of routines comprising:
- a routine configured to access a set of training images;
- and
- a routine configured to train an appearance model using the set of training images to learn an alignment score function that minimizes angles between gradient directions of the alignment score function and ideal travel directions to a desired alignment.
A manufacture comprising:
- one or more non-transitory, computer-readable media having executable instructions stored thereon, the executable instructions comprising:
- instructions adapted to access an image including a human face;
- and
- instructions adapted to align the human face using a discriminative face alignment model that includes a discriminative appearance model trained to estimate an alignment score function that minimizes angles between gradient directions of the alignment score function and vectors pointing in the direction of the maximum of the alignment score function.
Dependencies are inferred from the claim text.