Parameter estimation method, parameter estimation device and collation method
View Patent ↗A parameter estimation method for estimating a parameter by estimating maximum a posteriori probability for input data. Operation for the input data is expressed by an inner product for the input data and the inner product is replaced by a Kernel function. By using the calculation result of the kernel function, a parameter is estimated. The method includes a step (offline operation) for learning a correlation between a plurality of learning input data in which a parameter to be estimated is known and the parameter corresponding to each of the learning input data; and a step (online operation) for estimating the parameter for the estimating input data in which a parameter to be estimated is unknown, by using the learned correlation.
1. A collation method using a parameter estimation device, comprising:
learning means for learning a correlation between a plurality of learning input data in which a parameter to be estimated is known and a parameter corresponding to each of the learning input data; and
estimation means for estimating a parameter with respect to estimating input data in which a parameter to be estimated is unknown by using the learned correlation;
wherein the learning means including a learning kernel matrix calculation means for calculating a learning kernel matrix by using the plurality of learning input data in which a parameter to be estimated is known, and an eigenvalue calculation means for obtaining an eigenvalue and an eigenvector of an inverse matrix of the learning kernel matrix; and
the estimation means including an estimation kernel calculation means for calculating an estimation kernel by using the estimating input data in which a parameter to be estimated is unknown and the learning input data, and a parameter calculation means for calculating a parameter with respect to the estimating input data by using the learning kernel matrix, and the eigenvalue and the eigenvector of the inverse matrix of the learning kernel matrix;
wherein the estimating input data and the learning input data are an image including a face, and the parameter to be estimated is a coordinate of a feature of the face;
the method comprising:
estimating a coordinate of a feature of a face with respect to a face image to be collated by using the parameter estimation device;
then cutting out a face region based on the coordinate of the feature of the face; and
collating a face image registered in the face image database with the cut-out face region.
2. A collation method using a parameter estimation device, comprising:
learning means for learning a correlation between a plurality of learning input data in which a parameter to be estimated is known and a parameter corresponding to each of the learning input data; and
estimation means for estimating a parameter with respect to estimating input data in which a parameter to be estimated is unknown by using the learned correlation;
wherein the learning means including a learning kernel matrix calculation means for calculating a learning kernel matrix by using the plurality of learning input data in which a parameter to be estimated is known, and an eigenvalue calculation means for obtaining an eigenvalue and an eigenvector of an inverse matrix of the learning kernel matrix; and
the estimation means including an estimation kernel calculation means for calculating an estimation kernel by using the estimating input data in which a parameter to be estimated is unknown and the learning input data, and a parameter calculation means for calculating a parameter with respect to the estimating input data by using the learning kernel matrix, and the eigenvalue and the eigenvector of the inverse matrix of the learning kernel matrix;
wherein the estimating input data and the learning input data are an image including a face, and the parameter to be estimated is a coordinate of a feature of the face,
the method comprising:
repeating a face image normalization operation plural times, the operation including estimating a coordinate of a feature of a face by using the parameter estimation device and normalizing an image region based on the coordinate of the feature of the face;
then cutting out a face region based on the coordinate of the feature of the face; and
collating a face image registered in the face image database with the cut-out face region.
3. A collation method using a parameter estimation device, comprising:
learning means for learning a correlation between a plurality of learning input data in which a parameter to be estimated is known and a parameter corresponding to each of the learning input data; and
estimation means for estimating a parameter with respect to estimating input data in which a parameter to be estimated is unknown by using the learned correlation;
wherein the learning means including a learning kernel matrix calculation means for calculating a learning kernel matrix by using the plurality of learning input data in which a parameter to be estimated is known, and an eigenvalue calculation means for obtaining an eigenvalue and an eigenvector of an inverse matrix of the learning kernel matrix; and
the estimation means including an estimation kernel calculation means for calculating an estimation kernel by using the estimating input data in which a parameter to be estimated is unknown and the learning input data, and a parameter calculation means for calculating a parameter with respect to the estimating input data by using the learning kernel matrix, and the eigenvalue and the eigenvector of the inverse matrix of the learning kernel matrix;
wherein the estimating input data and the learning input data are an image including a face, and the parameter to be estimated is height information of the face image;
and a face image database,
the method comprising:
estimating height information with respect to a face image to be collated by using the parameter estimation device; and
then synthesizing a two-dimensional image seen from a same angle as an angle of a face image registered in the face image database based on the height information of the face image and carrying out a collation.
4. A collation method using a parameter estimation device, comprising:
learning means for learning a correlation between a plurality of learning input data in which a parameter to be estimated is known and a parameter corresponding to each of the learning input data; and
estimation means for estimating a parameter with respect to estimating input data in which a parameter to be estimated is unknown by using the learned correlation;
wherein the learning means including a learning kernel matrix calculation means for calculating a learning kernel matrix by using the plurality of learning input data in which a parameter to be estimated is known, and an eigenvalue calculation means for obtaining an eigenvalue and an eigenvector of an inverse matrix of the learning kernel matrix; and
the estimation means including an estimation kernel calculation means for calculating an estimation kernel by using the estimating input data in which a parameter to be estimated is unknown and the learning input data, and a parameter calculation means for calculating a parameter with respect to the estimating input data by using the learning kernel matrix, and the eigenvalue and the eigenvector of the inverse matrix of the learning kernel matrix;
wherein the estimating input data and the learning input data are an image including a face, and the parameter to be estimated is height information of the face image;
and a face image database,
the method comprising:
estimating height information with respect to a face image registered in the face image database by using the parameter estimation device;
then synthesizing a two-dimensional image seen from an angle that is different from an angle of the registered face image; and
additionally registering the synthesized two-dimensional image in the face image database.