Apparatus and method for extracting feature of image including object
AssigneesSamsung Electronics Co., Ltd., Northeastern University
InventorsSungjoo Suh, Yun Fu, Zhengming Ding, Chang Kyu Choi, Jae-Joon Han
Life of this patent
Estimating the term…
Abstract
At least one example embodiment discloses a method of converting a vector corresponding to an input image. The method includes receiving first-dimensional vector data associated with an input image, the input image including an object and converting the received first-dimensional vector data to second-dimensional vector data based on a projection matrix with an associated rank. A first dimension of the first-dimensional vector data is higher than a second dimension of the second-dimensional vector data.
Claims24
An object recognition method, the method comprising:
- receiving first-dimensional vector data associated with an input image, the input image including an object;
- converting the received first-dimensional vector data to second-dimensional vector data based on a projection matrix with an associated rank, the converting including,
- converting the received first dimensional vector data to the second-dimensional vector data such that the second-dimensional vector data is same as desired vector data, the desired vector data including a sparse error matrix and a product of basis vector data and a coefficient matrix, wherein a first dimension of the first-dimensional vector data is higher than a second dimension of the second-dimensional vector data;
- and
- performing object recognition based on the second-dimensional vector data.
A method of learning a projection matrix to convert a dimension of obtained vector data associated with an input image including an object, the method comprising:
- receiving learning vector data corresponding to each of a plurality of learning images;
- learning the projection matrix from the learning vector data, the projection matrix having an associated rank;
- and
- determining the projection matrix such that the obtained vector data is identical to desired vector data, the desired vector data including a sparse error matrix and a product of basis vector data and a coefficient matrix;
- wherein the determined projection matrix is used to perform an object recognition.
An object recognition method, the method comprising:
- extracting vector data associated with an input image from the input image;
- converting the extracted vector data to discriminative feature data based on a projection matrix, the projection matrix being based on first and second covariance matrices corresponding to learning vector data, the projection matrix having an associated rank;
- and
- performing object recognition based on the discriminative feature data.
An object recognition apparatus, the apparatus comprising:
- a memory storing computer-readable instructions;
- and
- at least one processor configured to execute the computer-readable instructions to,
- receive first-dimensional vector data associated with an input image, the input image including an object,
- convert the received first-dimensional vector data to second-dimensional vector data based on a projection matrix such that the second-dimensional vector data is same as desired vector data, the desired vector data including a sparse error matrix and a product of basis vector data and a coefficient matrix, and
- perform object recognition based on the second-dimensional vector data.
An apparatus for learning a projection matrix to convert a dimension of obtained vector data associated with an input image including an object, the apparatus comprising:
- a memory storing computer-readable instructions;
- and
- at least one processor configured to execute the computer-readable instructions to,
- receive learning vector data corresponding to each of a plurality of learning images,
- learn the projection matrix from the learning vector data, the projection matrix having an associated rank, and
- determine the projection matrix such that the learning vector data is identical to desired vector data, the desired vector data including a sparse error matrix and a product of basis vector data and a coefficient matrix,
- wherein the determined projection matrix is used to perform an object recognition.
Dependencies are inferred from the claim text.